<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Elucidata on Medium]]></title>
        <description><![CDATA[Stories by Elucidata on Medium]]></description>
        <link>https://medium.com/@elucidatainc?source=rss-71f407471440------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*IFM80hOsOVTK5U41_hZCqw.jpeg</url>
            <title>Stories by Elucidata on Medium</title>
            <link>https://medium.com/@elucidatainc?source=rss-71f407471440------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 11:37:32 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@elucidatainc/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[How PollyBERT Outperforms Large Language Models in BioNER]]></title>
            <link>https://medium.com/elucidata/how-pollybert-outperforms-large-language-models-in-bioner-9dc072542c9b?source=rss-71f407471440------2</link>
            <guid isPermaLink="false">https://medium.com/p/9dc072542c9b</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[dataset]]></category>
            <category><![CDATA[bert]]></category>
            <category><![CDATA[gpt-3]]></category>
            <dc:creator><![CDATA[Elucidata]]></dc:creator>
            <pubDate>Mon, 21 Nov 2022 13:57:29 GMT</pubDate>
            <atom:updated>2022-11-21T13:57:29.120Z</atom:updated>
            <content:encoded><![CDATA[<p>Large Language Models (or LLMs) are humongous models usually in the order of tens or hundreds of gigabytes which are trained on huge amounts of training data, which are generally in terabyte to the petabyte scale.</p><p>One of the most famous LLMs rights is <a href="https://en.wikipedia.org/wiki/GPT-3">GPT-3</a> with about 175 billion parameters and a 45TB big corpus. Some researchers calculated that if we were to train GPT-3 again using 1024 A100 GPUs it will still take us about 34 days [ <a href="https://arxiv.org/abs/2104.04473">1</a>]. It’s easy to see why a lot of individuals and organizations cannot afford to train an LLM.</p><blockquote><em>The silver lining with LLMs is that they are few-shot learners, which means that due to the sheer number of parameters and size of the training dataset they are able to perform tasks without being fine-tuned or with minimal tuning.</em></blockquote><p>Recently, we came across an <a href="https://towardsdatascience.com/a-quiet-shift-in-the-nlp-ecosystem-84672b8ec7af">article</a> that showed some interesting results using prompt design. The purpose of the article was to highlight the task-solving capabilities of LLMs which meet two conditions — the number of parameters in the model exceeds 100 billion and the amount of training exceeds 10 23FLOPS. This leads to the emergence of the model which gives it the capability to solve tasks without being specifically trained for them [3].</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*IwVlD4s_DPCqc4GO.png" /></figure><p>In the past, we have seen that vanilla natural language models tend to struggle in the biomedical domain. They usually need some fine-tuning or pre-training with biomedical data to give good results. Due to this our hypothesis is that the Quality of our dataset is more important than both the size of the model and the size of the corpus.</p><p>This intrigued us to explore LLMs with biomedical text data, specifically for Named Entity Recognition (NER) task and compare them against our BERT models which we use for NER on textual metadata from biomolecular datasets.</p><p>Before we jump to the experiments and numbers, let’s first define two more terms — <strong>domain-specificity</strong> and <strong>task-specificity</strong>, and what we mean by <strong>Quality of a dataset</strong>.</p><p><strong>Domain-specificity </strong>can be defined as the similarity between the domain of our dataset and the domain of the use case or problem statement we are developing a model for.</p><p><strong>Task-specificity </strong>can be defined as the similarity between the task that needs to be performed for our use case or problem statement and the task our dataset is built for.</p><p>According to our definition of the <strong>Quality of a dataset</strong>, it is a property of a dataset that will affect how well can a model learn from that dataset and this quality can be increased for a dataset by increasing domain-specificity and task-specificity.</p><p>Task-specificity should be an intuitive factor that can increase model performance. To learn patterns, a model needs a task to perform, and learn from the mistakes it makes while performing that task. While <strong>it is possible for a model to learn different tasks by using just one general task, empirically we have seen that fine-tuning for specific tasks yields better results</strong>.</p><p>Why domain specificity can help us train a better model, can also be explained using the distribution of our training and test sets. For Empirical Risk Minimization learning algorithms (which are almost all the supervised learning models we use), we need our training and test sets to be sampled from the same distribution. In the real world, after we serve a model we don’t have control over the test set but by <strong>increasing domain-specificity according to the use case, we can increase the similarity between the distribution of train and test examples</strong> since examples from the same domain will be similar. This becomes a bigger factor for biomedical data due to words that are unique to this domain.</p><h3>What Is Prompt-based Learning and How Do We Make Our Prompt?</h3><p>Prompt-based learning is an ML model training method in which we describe our task to perform with/without some examples. This description is then given to a pre-trained LLM for filling the missing information which is used as the output for the said task.</p><p>To create our prompt, we used the paper <a href="https://arxiv.org/pdf/2203.01543.pdf">QaNER</a> as a reference. In their paper, they suggested a Question Answering based NER framework and a method to convert NER problems to QA problems which we have used to create our prompts [ <a href="https://arxiv.org/abs/2203.01543">4</a>].</p><p>Along with methods from QaNER we also created 10 prompts and empirically selected the best prompts using a small test set. This was done to get the best prompt we can get so that the prompt is not a bottleneck and we can have an apples-to-apples comparison.</p><h3>How Do We Fine-tune PollyBERT?</h3><p>The task we want our BERT model to perform is NER. and for that, we use BERT in a token classification setting. In token classification, the model classifies each token as a certain entity or a regular token.</p><p>To fine-tune our model, we used a manually labelled dataset we created with Huggingface models and training APIs. The manually labelled corpus we used for fine-tuning PollyBERT was also a relatively small corpus with about 4k samples from <a href="https://www.ncbi.nlm.nih.gov/geo/">GEO</a> and only a few MBs in size.</p><h3>Experiments</h3><p>To test prompt design we picked GPT-J-6B and OPT-175B to represent two different sizes of LLMs, but their size is not the only difference between GPT-J-6B and OPT-175B. GPT-J-6B and OPT-175B were trained with the <a href="https://arxiv.org/abs/2101.00027">Pile Dataset</a> which is an 800GB big corpus, which includes natural language data from a lot of different sources. An important fact to note here is that OPT-175B was trained on a subset of Pile and this subset didn’t include PubMed Central and some other categories.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/806/0*YlwzCbGKxBrnf4iu.png" /></figure><p>For our BERT model, we used <a href="https://arxiv.org/pdf/2007.15779.pdf">PubMedBERT</a> weights for initializing a BERT model, PubMedBERT is a BERT-based model which has been pre-trained with a 21GB big PubMed abstract corpus. The PubMedBERT model was then fine-tuned for NER using a manually labelled training dataset. This dataset was created for extracting tissues from a given piece of text.</p><p>The task we chose for comparing these models was to perform NER and extract tissue entities from a piece of text. We are using the same prompt that was designed in the earlier section in a <strong>0-shot learning setting with OPT-175B and GPT-J-6B (due to API limitations)</strong>.</p><p>Since we had the capability to perform 2-shot learning with GPT-J-6B we also did that with the same test set to see if using 2-shot learning can help the model perform better.</p><p>To quantify performance for any model we are using both accuracy and F1-score as our metrics to measure performance for NER. One thing to note for these scores is that we are <strong>considering partial matches between the prediction and ground truth</strong>.</p><p>The test set we will be using is made using textual metadata from <a href="https://www.ncbi.nlm.nih.gov/geo/">GEO</a> which has been manually labelled by Subject Matter Experts (or SMEs). The test set is a perfectly balanced set with 100 examples containing tissue entities and 100 examples not containing any tissue entities.</p><h3>Results</h3><p>Now coming to numbers with PollyBERT we saw an accuracy of 87% and an F1 score of 0.84, GPT-J-6B gave us an accuracy of 51.7% and an F1 score of 0.515 and OPT-175B was not able to give any predictions hence accuracy and F1 score both stand at 0.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*r2MZmOdsIjTObmeK.png" /></figure><p>Plot to show as F1 score increases the model parameter count increases</p><p>For 2-shot learning with GPT-J-6B, we saw a further drop in performance, we got an accuracy of 47% and an F1 score of 0.38. Why this drop in performance happened is not clear to us, we suspect this might just be because of the test set.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/966/0*JC2fh6jOS_jX5JcO.png" /></figure><p>Comparing 2-shot learning with GPT-J-6B with PollyBERT</p><p>From these results, it’s clear that <strong>PollyBERT is the better performing model by a huge margin</strong> even though it has 10–1000 times lesser parameters and a significantly smaller training corpus than both GPT-J-6B and OPT-175B.</p><h3>Conclusion</h3><h4>Effect of Quality of the Dataset</h4><p>From our results, we have seen that PollyBERT performs much better than LLMs when it comes to NER for tissue entities. Also, we have a decrease in performance with an increase in the number of parameters. <strong>These results are completely opposite of what we expected given the condition of emergence was only met by OPT-175B</strong>.</p><p>According to us, this was due to the difference in the training dataset for these models. Biomedical data is very different from general language data, it has a lot of named entities that are only found in the biomedical domain. To identify these entities for NER, a model needs them in its corpus and since the occurrence of these entities will be negligible in a general language corpus it leads to poor performance.</p><p>If we recall the corpus for all these models, we used data from <strong>GEO for PollyBERT, GPT-J-6B had PubMed Central data and OPT-175B lacked any biomedical data</strong> while we were testing them on GEO. On top of that, while the dataset for <strong>PollyBERT was made for tissue NER, both GPT-J-6B and OPT-175B were trained using an autoregressive task</strong>.</p><p>If we look at domain-specificity and task-specificity here, the dataset for GPT-J-6B was more domain-specific than OPT-175B due to which GPT-J-6B performed better than OPT-175B. The dataset for PollyBERT was more task-specific and domain-specific (since they share the same source i.e., GEO) than GPT-J-6B which lead to a huge jump in performance.</p><h4>Manual Labelling vs Prompt Design</h4><p>After the quality of data, there is also a case to be made for using manual labelling instead of spending more time creating and experimenting with prompts, specifically for biomedical NER tasks.</p><p>Now, let’s assume we have an LLM which is trained a lot in biomedical data and is capable of performing a variety of tasks with prompt-based learning. If we go with prompt-based learning, we will decide on a framework for designing prompts like QaNER and then begin experimenting with different prompts on a test set. The problem with this method is that we are not guaranteed that we can design a prompt in the first place or the best possible one.</p><p>On the other hand, manual labelling has a defined process for getting labelled data with consistent results. Due to the reliability and consistency of manual labelling, it should be the preferred method of fine-tuning the language models for biomedical NER tasks.</p><blockquote><em>We are focussed on improving not just the accuracy but also the speed of the PollyBERT model. </em><a href="https://www.elucidata.io/blog/making-pollybert-run-10x-faster"><em>Read this blog to find out how some optimizations which were tried out recently made PollyBERT 10x faster!</em></a></blockquote><p><em>Originally published at </em><a href="https://www.elucidata.io/blog/how-pollybert-outperforms-large-language-models-in-bioner"><em>https://www.elucidata.io</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9dc072542c9b" width="1" height="1" alt=""><hr><p><a href="https://medium.com/elucidata/how-pollybert-outperforms-large-language-models-in-bioner-9dc072542c9b">How PollyBERT Outperforms Large Language Models in BioNER</a> was originally published in <a href="https://medium.com/elucidata">Elucidata</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Making PollyBERT Run 10x Faster]]></title>
            <link>https://medium.com/@elucidatainc/making-pollybert-run-10x-faster-83c2be6d03ac?source=rss-71f407471440------2</link>
            <guid isPermaLink="false">https://medium.com/p/83c2be6d03ac</guid>
            <category><![CDATA[bio-molecular]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[bioinformatics]]></category>
            <category><![CDATA[mlops]]></category>
            <category><![CDATA[drug-discovery]]></category>
            <dc:creator><![CDATA[Elucidata]]></dc:creator>
            <pubDate>Tue, 27 Sep 2022 08:03:12 GMT</pubDate>
            <atom:updated>2022-09-27T09:23:36.389Z</atom:updated>
            <content:encoded><![CDATA[<p>-Mukund Chaudhry</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kw1gDNxbZ5BusTpS3fV80A.png" /></figure><p><a href="https://www.elucidata.io/platform/features">Polly </a>is Elucidata’s data-centric MLOps platform which hosts millions of biomolecular datasets, and PollyBERT is our suite of transformer-based language models. The “datasets” we ingest usually come in the form of experimental measurements, accompanying textual metadata, and/or scientific papers describing the experiments’ results. The large number of datasets ingested makes it important to ensure an easy search and catalog function for our users. We do this by mapping datasets with metadata annotations. The annotation process, which would be extremely time-consuming if done manually, is aided by PollyBERT.‍</p><figure><img alt="Elucidata’s PollyBERT" src="https://cdn-images-1.medium.com/max/1024/0*9nFl9yBbrsDb1Lm2.png" /><figcaption>Annotations capture various aspects of the experimental design — sample source (organism, tissue, cell line), perturbation type (drug, genotype), disease condition, phenotype, etc. Users can use these annotations to perform complex searches to find datasets of their interest.‍</figcaption></figure><p>PollyBERT models are trained to perform information extraction tasks like sequence tagging, sequence classification, entity normalization, relation extraction, and slot-filling. They scan the text and papers and annotate them with <a href="https://www.elucidata.io/blog/data-centricity-is-the-future-of-biomedical-nlp">human-level accuracy</a>.‍</p><figure><img alt="Elucidata’s PollyBERT" src="https://cdn-images-1.medium.com/max/992/0*aCsT28Xx_geRVo9u.png" /></figure><h3>Time Is Money! Why We Optimized PollyBERT Models to Improve Throughput?</h3><p>If you’ve been following the news around NLP lately, you’ll know that <a href="https://arxiv.org/pdf/2004.08900.pdf">large language models are expensive</a> to train and perform inference on. The rate at which we’re ingesting new data into Polly only worsens this problem. To give some perspective, one forward pass through BERT takes 8–50ms on an AWS g4dn.xlarge instance (which comes with an NVIDIA T4 GPU). This means that running 1 BERT model on 1 million inputs would take ~ 9 hours. Since we have 12 different BERT models performing different tasks, it would take 5 days to run them sequentially on one machine. We often need to run these models on several million inputs every week, which can end up costing thousands of dollars every month. Last quarter, we decided to spend some time doing some optimization to help us increase the number of requests our models could handle. We summarized our problem statement:<em> Increase model inference throughput without lowering model accuracy or increasing costs.</em></p><p><strong><em>In this article,</em></strong> I describe some of the optimizations we tried and how effective they were at helping us achieve our goal. We ended up picking the methods that gave the biggest ROI.</p><blockquote>Note that some of these optimizations were specific to our use case, so they might not be universally applicable. Also, we used the BERT architecture for most of our experiments, so you can assume that these results are for BERT unless stated otherwise.</blockquote><p>Before we delve into the fancy model compression techniques, let’s talk about what it means to make models faster. When it comes to deep learning models, there are generally 2 variables that people optimize for before deployment.</p><p>These are:<strong><em><br>1. Size:</em></strong><br>This is the space taken up by the model’s weights, on disk and in RAM.<br>Vanilla BERT has a size of 400 MB.<strong><em><br>2. Inference Time:</em></strong><br>This is the time it takes for one prediction.<br>For our corpus, vanilla BERT took 30ms (avg) on GPU and 430ms (avg) on CPU.</p><p>Model size tends to matter more when deployed on edge devices and mobile phones. In that case, the size can directly impact user experience since it affects download time and space taken up on the device. If your model runs on servers, the model size can affect how quickly you can spin up new instances when the load increases. For us, neither of these was a concern. Our models run server-side, not on edge devices, and the load doesn’t change fast enough to warrant a need for quick spin-up times. We ended up not taking model size into account when deciding on which optimizations to go with.</p><p>We were mainly concerned with how many inputs our servers could handle per second (throughput), which dictates how many datasets we can annotate. <strong><em>Since inference time directly impacts throughput, that’s the quantity we’re implicitly optimizing for.</em></strong>‍</p><p><strong>Controlling for Sequence Length and Hardware</strong></p><p>Different factors can affect BERT’s inference time. This includes batch size, sequence length, choice of deep-learning framework, and hardware. We used Pytorch and a batch size of 1 for all our experiments, a g4dn.xlarge instance as our GPU machine and a laptop with Intel i5–7300U as our CPU machine. Furthermore, we fixed the corpus used to test different model variants to control for sequence length, ensuring it resembled production workloads. For our corpus, vanilla BERT took 30ms per input (avg) on GPU and 430ms per input (avg) on CPU.</p><h3>Quantizing Model Parameters from float32 to int8</h3><p>Model parameters and activations are typically stored as float32. BERT has 110 million float32 parameters, which take up 410MB of memory. The idea behind quantization is to use float16 or int8 to store these parameters, thereby reducing the computation time and the model size.</p><p>There are different ways to do this quantization. One of them is called <em>dynamic quantization</em>. Here the mapping from float32 to int8 happens <em>after training</em>. The model parameters are converted to int8 by rounding them prior to inference. This part is straightforward but not enough. If we leave it there, the errors due to quantization can propagate through the layers and become increasingly big. These errors are reduced by having the <em>activations</em> undergo an additional <a href="https://en.wikipedia.org/wiki/Affine_transformation">affine transformation</a> (y = Ax + b). This transformation is done in a way that ensures that the model’s activations deviate as little as possible from their original (pre-quantization) values. The <em>A</em> and <em>b</em> for the transformation are learned dynamically during inference (see <a href="https://leimao.github.io/article/Neural-Networks-Quantization/">here</a> for details). Also, not all operations are performed using int8. Some use int16 or int32 to avoid issues like integer overflow.</p><p>Here’s a diagram to show what the weights look like before and after quantization. You can clearly see the decrease in floating-point precision.</p><figure><img alt="Quantization of Model Parameters from float32 to int8" src="https://cdn-images-1.medium.com/max/963/0*Nf6S2bG-fqtCaLlX.png" /><figcaption>Histogram of 2.5k randomly sampled parameters, (left) before quantization and (right) after quantization.‍</figcaption></figure><p>We got a 60% reduction in model size and a 1.6x speedup on the CPU using dynamic quantization. Unfortunately, if you’re using Pytorch (which we do), quantization benefits are <a href="https://pytorch.org/docs/stable/quantization.html#natively-supported-backends">not available on GPU</a>.</p><p>There are two other well-established ways of quantizing your network, namely static quantization and quantization-aware training, each with its advantages. We stuck to dynamic quantization because it is super easy to implement. Using Pytorch, it can be done in a few lines and doesn’t differ too much from the other two in terms of inference time reduction.</p><h3>Parameter Pruning</h3><p>As the name suggests, <a href="https://towardsdatascience.com/model-compression-via-pruning-ac9b730a7c7b">pruning </a>involves removing (or masking) model parameters that don’t contribute much to accuracy but use a lot of computational resources. Pruning can be done by zeroing parameters that are already close to zero (<em>unstructured pruning</em>) or in a more systematic way by removing entire layers and reshaping weight matrices and, in some cases, re-training the network afterward (structured pruning).</p><p>Pruning is an interesting technique in itself. It can be used both as a way of making your model run faster and as a means to explore what your model is learning.</p><p>Structured pruning can be quite effective at speeding up inference, but it is also a fairly elaborate process involving a lot of trial and error. Since we were short on time, we decided to table this for later.</p><p>We did try out unstructured pruning. We could zero as many as 30% of our model’s parameters without seeing any drop in F1. Unfortunately, this did not give us any inference speedup. The reason for this is that, while unstructured pruning increases the sparsity of your network, most deep-learning frameworks (including Pytorch) are not currently designed to take advantage of this sparsity to speed up computation (see <a href="https://discuss.pytorch.org/t/weight-pruning-on-bert/83429/2">here</a>). They perform the same number of operations regardless of how sparse your network is. Unstructured pruning can help in reducing model size, but only if you store the network weights using some sparse vector format.</p><h3>Using Knowledge Distillation (KD) to Train <a href="https://huggingface.co/docs/transformers/model_doc/distilbert">Distilbert</a></h3><p><a href="https://en.wikipedia.org/wiki/Knowledge_distillation">KD</a> is a way of training a smaller student model from a larger teacher model. The key idea here is that we teach the student to mimic the teacher’s output distribution (soft labels) instead of training it directly for a given task. Empirical results have shown that training the student in this manner is more effective. There are a few different ways BERT can be made “smaller,” but decreasing the number of layers is the most effective.</p><p>For our experiments, we used the <a href="https://huggingface.co/docs/transformers/model_doc/distilbert">distilbert</a> architecture. We trained it using a fine-tuned BERT as the teacher. Distilbert gave us a speedup of 1.5x on GPU (and 2.4x on CPU) but had a 5-point lower F1 score. We suspect the lower F1 is because Distilbert was pre-trained on general language text. Our best-performing models are initialized from BioBERT and PubmedBERT, both of which are pre-trained on biomedical text. There doesn’t seem to be any distil-<em>bio</em>-bert available in the public domain. Perhaps, in the near future, we’ll invest some time in training one ourselves.</p><p><em>Having read about pruning and KD, you might be wondering, if we can reach the same accuracy using a smaller network, why train the bigger network in the first place? Why not just train the smaller network directly?<br>The short answer is, bigger, over-parameterized models converge more easily to an optimal solution.<br>The </em><a href="http://mitchgordon.me/machine/learning/2020/01/13/do-we-really-need-model-compression.html"><em>Long answer</em></a><em> makes for an interesting read.</em></p><p>To read the entire blog and get an in-depth understanding, <a href="https://bit.ly/3SBHr3T">click here</a>!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=83c2be6d03ac" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Dataset of the week — GSE102956]]></title>
            <link>https://medium.com/elucidata/dataset-of-the-week-gse102956-2dcd03d41c0e?source=rss-71f407471440------2</link>
            <guid isPermaLink="false">https://medium.com/p/2dcd03d41c0e</guid>
            <category><![CDATA[homo-sapiens]]></category>
            <category><![CDATA[drug-targets]]></category>
            <category><![CDATA[alzheimers-disease]]></category>
            <category><![CDATA[omics]]></category>
            <category><![CDATA[crispr-cas9]]></category>
            <dc:creator><![CDATA[Elucidata]]></dc:creator>
            <pubDate>Thu, 27 May 2021 18:48:58 GMT</pubDate>
            <atom:updated>2021-05-27T18:50:11.188Z</atom:updated>
            <content:encoded><![CDATA[<h3>Dataset of the week -GSE102956</h3><p>by Shruti Mehrotra</p><p><em>Our ‘Dataset of the Week’ series features publicly available omics datasets of scientific value, intending to promote data sharing and reuse.</em></p><p><strong>Publication</strong>: Neuron, 2018.</p><p>This week, we feature RNA-seq profiles of three iPSC-derived brain cell types from the publication ‘<a href="https://www.cell.com/neuron/fulltext/S0896-6273(18)30380-5?_returnURL=https%3A%2F%2Flinkinghub.elsevier.com%2Fretrieve%2Fpii%2FS0896627318303805%3Fshowall%3Dtrue">APOE4 Causes Widespread Molecular and Cellular Alterations Associated with Alzheimer’s Disease Phenotypes in Human iPSC-Derived Brain Cell Types</a>’(1).</p><p><strong>Clinical Problem</strong>:</p><p>The apolipoprotein E4 (APOE4) has gained much attention in recent years as the most significant genetic risk factor for sporadic Alzheimer’s disease (sAD)(2). Conformational changes in the secretion of APOE4 proteins by astrocytes, neurons, and microglia lead to various pathological effects unique to sAD. However, prior studies have failed to evaluate the significance of APOE4 expression in brain cells in the context of humans. Researchers have used the CRISPR/Cas9 technology to produce human-induced pluripotent stem cells (hiPSCs) mimicking AD phenotypes in 3 types of brain cells to bridge this gap. The paper also explores the physiopathological effects of removing the sAD causing factor from these stem cells.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/375/1*pLY3KPARt_5490NTKp8eRA.png" /><figcaption>Graphical Abstract</figcaption></figure><p><strong>Experimental Design for the datasets</strong>:</p><p>Parental APOE3 cells from an unaffected subject were used to generate APOE4 iPSCs using the CRISPR/Cas9 method. These cells were further differentiated into neurons, astrocytes, and microglia-like cells in replicates of 3. Further, exome-seq data (76-bp paired-end) were generated for each type of brain cells. Gene expressions underlying the cellularization were detailed using these RNA-seq datasets.</p><p><strong>Key findings of the publication</strong>:</p><ul><li>APOE4 caused accelerated maturation of neurons as well as upregulated genes related to neuronal differentiation.</li><li>Significant downregulation of the BDNF gene associated with the synaptic function was noted. This is consistent with the cognitive impairment phenotype associated with the BDNF gene in previous studies (3,4).</li><li>A cell-specific reduction in APOE protein transcription was noted in astrocytes indicating that the APOE4 variant can impact its own transcription.</li><li>The APOE4 variant regulates the expression of both metabolism and transport of lipids such as<br>cholesterol, with potential effects on each of the three brain cell types.</li><li>APOE4 glial cells were observed to be impaired in their ability to uptake and eliminate extracellular beta- amyloids (Aβ) peptides.</li></ul><p><strong>Significance of the dataset</strong>:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/695/1*SMWhFfwNZ8Ov0Nb9KxUzzg.png" /><figcaption>A PCA plot depicting clusters of iPScs from a normal individual (Purple), iPSCs derived from neurons (Olive), and iPSCs derived from astrocytes(Brown). Plot generated on Polly.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/671/1*VFmgdeO_jN8vnj7cjM_gJA.png" /><figcaption>Volcano plot showing the statistical significance of several genes, including OSMR (Oncostatin M Receptor). OSMR is expressed in the glial cells and is involved with the formation of amyloids in patients with neurodegenerative diseases. Plot generated on Polly.</figcaption></figure><p>Links: <a href="https://www.cell.com/neuron/fulltext/S0896-6273(18)30380-5?_returnURL=https%3A%2F%2Flinkinghub.elsevier.com%2Fretrieve%2Fpii%2FS0896627318303805%3Fshowall%3Dtrue">Full Article</a> | <a href="https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE102956">Dataset: GSE102956</a> | <a href="https://pubmed.ncbi.nlm.nih.gov/29861287/">PMID</a> | <a href="https://linkinghub.elsevier.com/retrieve/pii/S0896627318303805">DOI</a> | <a href="https://omixwiki.elucidata.io/507/gse102956/apoe4-causes-widespread-molecula">OmixWiki</a> | <a href="https://elucidata.io/discover/">Polly</a></p><p><strong>References:</strong></p><ol><li>Lin YT., et al. APOE4 Causes Widespread Molecular and Cellular Alterations Associated with Alzheimer’s Disease Phenotypes in Human iPSC-Derived Brain Cell Types. <em>Neuron</em>. vol. 98(6): 1141–1154 (2018).</li><li>Cataldo, A., et al. Endocytic Pathway Abnormalities Precede Amyloid β Deposition in Sporadic Alzheimer’s Disease and Down Syndrome: Differential Effects of APOE Genotype and Presenilin Mutations. <em>The American Journal of Pathology</em>, vol. 157(1): 277–286 (2000).</li><li>Zhao J., et al. APOE4 exacerbates synapse loss and neurodegeneration in Alzheimer’s disease patient iPSC-derived cerebral organoids. <em>Nature Communications</em>. vol. 11(1): 5540 (2020).</li><li>Zhang L., et al. Brain-Derived Neurotrophic Factor Ameliorates Learning Deficits in a Rat Model of Alzheimer’s Disease Induced by Aβ1–42. <em>PLoS ONE</em>. vol. 10(4): e0122415 (2015).</li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2dcd03d41c0e" width="1" height="1" alt=""><hr><p><a href="https://medium.com/elucidata/dataset-of-the-week-gse102956-2dcd03d41c0e">Dataset of the week — GSE102956</a> was originally published in <a href="https://medium.com/elucidata">Elucidata</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Hiring Engineering Leaders for Scaling Bioinformatics 101]]></title>
            <link>https://medium.com/elucidata/hiring-engineering-leaders-for-scaling-bioinformatics-101-10a91b4c5fa2?source=rss-71f407471440------2</link>
            <guid isPermaLink="false">https://medium.com/p/10a91b4c5fa2</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[hiring]]></category>
            <category><![CDATA[bioinformatics]]></category>
            <category><![CDATA[engineering]]></category>
            <category><![CDATA[full-stack]]></category>
            <dc:creator><![CDATA[Elucidata]]></dc:creator>
            <pubDate>Mon, 24 May 2021 12:01:39 GMT</pubDate>
            <atom:updated>2021-05-24T12:01:39.592Z</atom:updated>
            <content:encoded><![CDATA[<p>by Swetabh Pathak</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ZsNVUyiZaWGbW3x6.png" /></figure><p>We, at <a href="https://elucidata.io/">Elucidata</a>, work with many small to mid-size biopharma companies that use large amounts of high throughput data. They generate this data in-house or through CRO partners. Apart from this, they are also looking at multiple public data sources — data repositories, databases and papers.</p><p>For a while, the teams are composed of just bioinformatics scientists. If you are lucky or have hired well — your team can make and deploy production-ready pipelines on their own. But at some point, that isn’t enough. You need a way to link your data store to notebooks, connect interactive environments to production-ready pipelines and build dashboards. Like lots of dashboards!</p><p>This is when many of our partners identify the need for an engineering leader. You need someone who can think of software maintenance, building scalable infrastructure, enforcing code practices and more.</p><p>As pharma companies reorient themselves to tackle the <a href="https://www.nature.com/articles/498255a">biological big data revolution</a>, bioinformatics leaders often find themselves tasked with filling technical/engineering positions within the organization to help meet their team’s growing computational needs. Hiring for a position that you have no background or training in can be very challenging. Ask me! I have hired for many roles where I couldn’t pass the most basic first round.</p><p>If you’re a bioinformatics scientist looking to hire an engineering leader for your team, here’s a handy guide to help you get started:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*E-Ej5vtJmMR1glWm.png" /></figure><p><strong>1. Identify the technical gap.</strong></p><p>The first step of the hiring process is to identify the gaps in knowledge on your team. Different roles will require different areas of expertise (see infographic). Narrowing down on the exact technical requirements will help you in your search for potential candidates and simplify the selection process. For instance, if your team is required to build a large number of dashboards, you can anticipate a big front-end gap. On the other hand, if your team deals with a high throughput of next-generation sequencing (NGS) data, you probably want to hire someone with significant experience in cloud computing services like AWS, GCP or similar platforms of your choice.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ucnVQzVK8tcJysm6.png" /></figure><p><strong>2. Remember that you are a biopharma company, not a tech start-up.</strong></p><p>If you’re hiring an engineering leader at a tech company, you’re looking for someone who can build scalable technology and grow and manage teams of developers. On the contrary, at a biopharma company, the focus is on enabling bioinformaticians/bench-scientists to be more efficient. So advice which is excellent for tech companies, such as <a href="https://a16z.com/2017/05/26/hiring-vp-engineering-why-what/">this post</a> from a famous technology venture firm, might not necessarily work for you. The person you are looking for is more of an individual contributor who can also play a bit of the product manager role. Tech companies have specialists to do product/project management. In your case, it’s more likely that the head of engineering is also talking to users to understand requirements and translate them into products. This is NOT what a traditional head of engineering at a tech start-up would do. They would focus a lot more on hiring, building engineering culture and be answerable for the final engineering delivery.</p><p><strong>3. Know when to buy versus build.</strong></p><p>Once you’ve figured out what kind of technical expertise your team needs, you may be able to work with contract tech professionals instead of hiring a full-time employee. This is a common trend in tech; <a href="https://www.forbes.com/sites/johanmoreno/2019/05/31/google-follows-a-growing-workplace-trend-hiring-more-contractors-than-employees/?sh=4f96ef15447f">Google reportedly employs 102,000 temporary/contract workers worldwide</a>. The question is when to opt for a contractor. The most common reason is when you require someone with a niche specialization for a fairly high-level need. Hiring a subject matter expert for a short-term project, such as a cloud migration specialist to get your cloud infrastructure up and running, would serve you better economically, as opposed to hiring a full-time employee. Outsourcing could also be a good long-term strategy when the requirement is for a small team that doesn’t have to be scaled.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*GhKtd3VIcEDiOEIM.png" /></figure><p><strong>4. Hire someone comfortable with working with contractors or outsourcing companies.</strong></p><p>Leading up from the previous point, It’s unlikely that you can hire for all the diverse skills that a bioinformatics platform needs. Budgets won’t allow it. In that case, your leadership might need to more open to hiring contractors or outsourcing companies. However, If your engineering leader wants to do everything in-house, you can run into problems. In tech companies, engineering leaders primarily focus on growing their teams and managing them. On the other hand, biopharma companies would require them to work with a limited team of engineers and be more involved in project and stakeholder management. In this scenario, you’re better off hiring a candidate who is amenable to leading a fairly small team of engineers while working with and managing contractors as required.</p><p><strong>5. Provide context to attract and retain the right tech talent.</strong></p><p>Engineers and developers who lack pharma industry experience may not understand how their work will contribute to the team and its growth. This may <a href="https://www.oracle.com/us/media1/talent-retention-6-best-practices-1676595.pdf">make it hard to attract and retain talent.</a> It’s important to provide this context through a well-crafted job description that talks about specific technical requirements, day-to-day responsibilities as well as how the potential hire’s contributions will fit into the larger vision of the company. The better you are at telling your story and specifying what you want, the easier it will be to find the right candidate.</p><p><strong>6. Look for someone who is on top of biopharma SaaS</strong></p><p>Over the past decade, the intersection of technology and biomedical research <a href="https://www.nature.com/articles/d41586-018-05267-x">has considerably accelerated the process of drug discovery</a>. As Pieter Peeters of Janssen Research &amp; Development put it, “AI is affecting every aspect of the pharmaceutical industry, from the very early drug discovery aspects, all the way through to how we do clinical development and bring medicines to patients.” There are a plethora of SaaS products on the market today that cater specifically to the data-intensive needs of biopharma companies. This ranges from access to ML-ready biomedical data (<a href="https://elucidata.io/">Elucidata</a>) to cloud-based services for processing (<a href="https://databricks.com/">Databricks</a>) and deploying machine learning models (<a href="https://aws.amazon.com/sagemaker/">AWS SageMaker</a>) on massive quantities of high-throughput data. In order to be a good fit, an ideal candidate would be someone who is excited about harnessing technology to solve problems in the pharma sector and keeps up with the latest technology trends in this space.</p><p><strong>7. Leadership qualities are as important as technical capabilities.</strong></p><p>While it’s a given that a technical lead should have superior coding skills, you should also be looking for someone with prior project and people management experience, the ability to serve as a good mentor, and has great communication and conflict resolution skills. Another underrated quality is the ability to delegate. It may be tempting for a tech lead to tackle every challenging problem hands-on, but this might leave them with limited bandwidth for their other responsibilities. Look for someone who not only surpasses the technical criteria but is also open to delegating important tasks to other team members without micromanaging them.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*KMQ_u05Voas2QfM2.png" /></figure><p><strong>8. Involve someone with prior experience in hiring for similar technical roles.</strong></p><p>It’s never a bad idea to ask for help when you need it. Talk to your friends, family, mentors, etc. to find someone who has prior experience hiring for similar positions. Get them involved in the process at some level. They will have a better sense of what to look for in terms of technical abilities as well as cultural fit for your team. If you have questions or would like to know more about hiring engineering leaders for a pharmaceutical company, feel free to reach out to me on <a href="https://www.linkedin.com/in/swetabh-pathak-b4154558/?originalSubdomain=in">Linkedin</a> or via email at swetabh.pathak@elucidata.io.</p><p>— <a href="https://www.linkedin.com/in/swetabh-pathak-b4154558/?originalSubdomain=in">Swetabh Pathak</a> is a co-founder of <a href="https://elucidata.io/">Elucidata</a>, the world’s largest repository of ML-ready biomedical multi-omics data.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=10a91b4c5fa2" width="1" height="1" alt=""><hr><p><a href="https://medium.com/elucidata/hiring-engineering-leaders-for-scaling-bioinformatics-101-10a91b4c5fa2">Hiring Engineering Leaders for Scaling Bioinformatics 101</a> was originally published in <a href="https://medium.com/elucidata">Elucidata</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Dataset of the week — GSE70955]]></title>
            <link>https://medium.com/elucidata/dataset-of-the-week-gse70955-28fe91ac31f1?source=rss-71f407471440------2</link>
            <guid isPermaLink="false">https://medium.com/p/28fe91ac31f1</guid>
            <category><![CDATA[bioinformatics]]></category>
            <category><![CDATA[muscular-dystrophy]]></category>
            <category><![CDATA[microarray-techniques]]></category>
            <category><![CDATA[omics]]></category>
            <category><![CDATA[drug-discovery]]></category>
            <dc:creator><![CDATA[Elucidata]]></dc:creator>
            <pubDate>Fri, 21 May 2021 20:23:26 GMT</pubDate>
            <atom:updated>2021-05-21T20:23:26.943Z</atom:updated>
            <content:encoded><![CDATA[<h3>Dataset of the week — GSE70955</h3><p>by Shruti Mehrotra</p><p><em>Our ‘Dataset of the Week’ series features publicly available omics datasets of scientific value, intending to promote data sharing and reuse.</em></p><p><strong>Publication</strong>: Cell Reports, 2016.</p><p>This week, we feature a microarray gene expression profile generated on an Affymetrix PrimeView human oligonucleotide array from the publication titled ‘<a href="https://www.cell.com/cell-reports/fulltext/S2211-1247(16)30584-8?_returnURL=https%3A%2F%2Flinkinghub.elsevier.com%2Fretrieve%2Fpii%2FS2211124716305848%3Fshowall%3Dtrue">Concordant but Varied Phenotypes among Duchenne Muscular Dystrophy Patient-Specific Myoblasts Derived using a Human iPSC-Based Model</a>’(1).</p><p><strong>Clinical Problem</strong>:</p><p>Duchenne Muscular Dystrophy (DMD) is characterized by deteriorated muscle functioning caused by degeneration of myocytes (muscle cells). This genetic disorder is caused by a dysfunction of the DMD gene that regulates the production of dystrophin, contributing to the reinforcement of muscle fibers. Although previous studies have documented histological pathologies and molecular mechanisms of DMD in animal models, human cell DMD models have been rarely assessed(2,3). Researchers have used hiPSCs to demonstrate an in-vitro DMD model in humans that exhibited patient-specific dystrophin mutations.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/375/1*Hb72D6T0Bc00dpQbc6n9JA.png" /><figcaption>Graphical Abstract</figcaption></figure><p><strong>Experimental Design</strong>:</p><p>Human embryonic stem cells were differentiated into skeletal muscle cells and compared to non-differentiated cells in triplicates. Microarray gene expression profiles were generated for the confirmation of differentiated cell characteristics. Global mRNA profiles were used to detail the gene expression underlying cellularization and to identify up-regulated genes involved in this process. Furthermore, microarray datasets were used for comparing control and disease gene expression profiles.</p><p><strong>Key findings of the publication</strong>:</p><ul><li>DMD myoblasts cultured in this experiment displayed patient-specific phenotypes — including altered transcriptional profiles, aberrant intracellular signaling, and defective myotube formation.</li><li>Using genetic correction and pharmacological “dual-SMAD” inhibition, the DMD-hiPSC-derived myoblasts formed “rescued” myotubes.</li><li>Affected myoblasts were observed to have increased BMP4 and TGFB signaling levels with aberrant expression of interleukins 6 and 8 and collagen 3 forming genes.</li><li>SPP1 (Osteopontin), BMP4, CASPASE1, BMP2, and TGFβ3 were reported to be the most positively correlated genes in the case of DMD myoblasts. SPP1 has been shown to be a potent genetic modifier of disease severity in DMD in previous studies(4).</li></ul><p><strong>Significance of the dataset</strong>:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/449/1*HGyNWBq1Pg2SH7GZDF4nWA.png" /><figcaption>PCA plot depicting clusters of normal embryonic stem cells (olive) and DMD-hiPSC-derived cells (purple). The plot is generated on Polly.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/654/1*q1kaSebBRTqqx6lBzcZY7A.png" /><figcaption>Volcano plot showing the statistical significance of several genes including MYOZ1 (Myozenin 1) which is involved with muscle strength. Plot generated on Polly.</figcaption></figure><p><strong>Links</strong>: <a href="https://www.cell.com/cell-reports/fulltext/S2211-1247(16)30584-8?_returnURL=https%3A%2F%2Flinkinghub.elsevier.com%2Fretrieve%2Fpii%2FS2211124716305848%3Fshowall%3Dtrue">Full Article</a> | <a href="https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE70955">Dataset: GSE70955</a> | <a href="https://pubmed.ncbi.nlm.nih.gov/27239027/">PMID</a> | <a href="https://www.cell.com/cell-reports/fulltext/S2211-1247(16)30584-8?_returnURL=https%3A%2F%2Flinkinghub.elsevier.com%2Fretrieve%2Fpii%2FS2211124716305848%3Fshowall%3Dtrue">DOI</a> | <a href="https://omixwiki.elucidata.io/506/gse70955/human-ipsc-based-model-for-duchenhttps:">OmixWiki </a>| <a href="https://elucidata.io/discover/">Polly</a></p><p><strong>References:</strong></p><ol><li>Choi IY., et al. Concordant but Varied Phenotypes among Duchenne Muscular Dystrophy Patient-Specific Myoblasts Derived using a Human iPSC-Based Model. <em>Cell Rep</em> vol. 15, 2301–12 (2016).</li><li>Barberi, T., Bradbury M., et al. Derivation of engraftable skeletal myoblasts from human embryonic stem cells. <em>Nat. Med.</em>1vol. 3, 642–648 (2007).</li><li>Li, H.L., Fujimoto, N., et al. Precise correction of the dystrophin gene in Duchenne muscular dystrophy patient induced pluripotent stem cells by TALEN and CRISPR-Cas9. Stem <em>Cell Reports</em> vol. 4, 143–154 (2015).</li><li>Pegoraro, E., Hoffman, E.P., et al. Cooperative International Neuromuscular Research Group SPP1 genotype is a determinant of disease severity in Duchenne muscular dystrophy. <em>Neurology</em> vol. 76, 219–226 (2011).</li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=28fe91ac31f1" width="1" height="1" alt=""><hr><p><a href="https://medium.com/elucidata/dataset-of-the-week-gse70955-28fe91ac31f1">Dataset of the week — GSE70955</a> was originally published in <a href="https://medium.com/elucidata">Elucidata</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Achieving FAIR Data]]></title>
            <link>https://medium.com/elucidata/achieving-fair-data-137c8293ef25?source=rss-71f407471440------2</link>
            <guid isPermaLink="false">https://medium.com/p/137c8293ef25</guid>
            <category><![CDATA[big-data]]></category>
            <category><![CDATA[polly]]></category>
            <category><![CDATA[fair-data]]></category>
            <category><![CDATA[fair]]></category>
            <category><![CDATA[reproducibility]]></category>
            <dc:creator><![CDATA[Elucidata]]></dc:creator>
            <pubDate>Mon, 05 Oct 2020 18:53:29 GMT</pubDate>
            <atom:updated>2020-10-05T18:53:29.106Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-4xHWAHSUDdYJOA5pdyyOg.png" /></figure><h3>Why FAIR?</h3><p>Transparency, openness, and reproducibility are readily recognized as vital features of science. One might expect that these valued features would be routine in daily practice, but growing evidence suggests that this is not the reality. The present culture that encourages innovation can undermine practices that support verification. Traditional practices in the view of scholarly communication are where data only serves as a by-product of research, are a tool to article publications, and are forgotten once they have served their purpose. <br>Moreover, there are no universally accepted measures that provide norms and guidelines that enable open practices in research. Most research institutions and labs have their own guidelines that pave the road to open science with obstacles.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/677/1*MStDbfdIfoPbaoEBvbCirQ.png" /></figure><p>Wilkinson <em>et. al.</em> 2016 published an open article that coined the abbreviation called FAIR i.e. FINDABLE, ACCESSIBLE, INTEROPERABLE, and REUSABLE. Making data FAIR is about data being as open as possible and as closed as necessary. The principles provide a set of guidelines for publication of digital resources such as datasets, code, workflows, and research objects, in a manner that makes them FAIR.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nyI9-4hCKtYmVNXyLJaFhw.jpeg" /></figure><p>With the rising awareness of the FAIR guiding principles, it is safe to say that these guidelines are universally accepted across domains and institutions, even though, there exists no universally accepted set of metrics that are applicable to all digital objects. The reasons being, first, FAIR guidelines provide a set of guiding principles that are more subjective in nature. Second, the stakeholders, data stewards have their own interpretation of a FAIR digital object.</p><p>There are numerous reasons as to how making data FAIR is beneficial to an organisation, some of which are as follows</p><ul><li>Time conservation for data collection</li><li>Increased citations</li><li>Encourages more funding opportunities</li><li>Prevents data loss</li><li>Benefits stakeholders</li><li>Data reuse for innovation in private sectors</li></ul><h3>What does FAIR mean to us?</h3><p>Metadata are crucial ingredients to perform the evaluation. Rich and descriptive metadata that is machine-actionable can be utilised are key tools to evaluate and answer questions that are asked about the data through a well-defined set of metrics. These measurable indicators give a clear idea about the extent of information that is available and steps that can be taken to maximise the FAIRness of data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/737/1*1m_ZwFCTGyIihYZ4HzHlzA.jpeg" /></figure><p>Making data FAIR is a continuous process, especially in the field of biomedical science due to rapid advancement in data mining and curation techniques. It is of utmost importance to provide datasets that follow domain-relevant standards. In the era of big data, it has also been realised that making data FAIR also means to reduce human intervention as much as possible, in order to ensure robust data analysis. “Timeliness” of biomedical data sharing is a crucial factor, especially during public health emergencies, in order to ensure that research communities can collaborate effectively and advance the speed of response and discovery. Hence, data (especially, data that is public) should be made FAIR as early as possible.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=137c8293ef25" width="1" height="1" alt=""><hr><p><a href="https://medium.com/elucidata/achieving-fair-data-137c8293ef25">Achieving FAIR Data</a> was originally published in <a href="https://medium.com/elucidata">Elucidata</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The №1 Thermo Scientific Compound Discoverer Node — Elucidata]]></title>
            <link>https://medium.com/elucidata/the-1-thermo-scientific-compound-discoverer-node-elucidata-2133433aabda?source=rss-71f407471440------2</link>
            <guid isPermaLink="false">https://medium.com/p/2133433aabda</guid>
            <category><![CDATA[scaling]]></category>
            <category><![CDATA[reproducibility]]></category>
            <category><![CDATA[thermo-fisher-scientific]]></category>
            <category><![CDATA[visualization]]></category>
            <category><![CDATA[metadata]]></category>
            <dc:creator><![CDATA[Elucidata]]></dc:creator>
            <pubDate>Fri, 31 Jul 2020 12:26:01 GMT</pubDate>
            <atom:updated>2020-07-31T12:26:01.175Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="The №1 Thermo Scientific Compound Discoverer Node — Elucidata" src="https://cdn-images-1.medium.com/max/1024/1*Ve29gFn6bqu4DvpG5ZmaCw.png" /></figure><h3>The №1 Thermo Scientific Compound Discoverer Node — Elucidata</h3><p>From peaks to absolute concentrations within minutes for 100s of metabolites</p><p>By <a href="https://www.linkedin.com/in/mrjains/">Vipul Jain</a></p><h3>About Polly™ QuantFit</h3><p><a href="https://elucidata.io/platform">Polly™ QuantFit</a> is a calibration tool used for the absolute quantification of metabolites. For the calculation of absolute concentration, the experiment is done with some standard samples of known concentrations.</p><figure><img alt="The №1 Thermo Scientific Compound Discoverer Node — Elucidata" src="https://cdn-images-1.medium.com/max/1024/0*si-MrJ5zoJGtb69d.png" /></figure><p>1. Intensities are detected in <a href="https://resources.elucidata.io/elmaven">El-MAVEN</a> across all metabolites for all samples</p><p>2. This workflow can further be customized as per experimental designs to perform normalization</p><p>3. Get mathematical mappings from intensities to concentrations for the standard samples, which are used to calculate concentrations of experimental intensities in Polly™ QuantFit</p><p>4. Generate customized visualizations with Ipythons or a concentration plot used for downstream processing</p><figure><img alt="The №1 Thermo Scientific Compound Discoverer Node — Elucidata" src="https://cdn-images-1.medium.com/max/1024/0*bkT4aiopA38eXTRP.png" /></figure><h3>Polly™ QuantFit Features</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-ozvuIy6u98zEOIaCePHAQ.png" /></figure><h3>About Thermo Scientific™ Compound Discoverer™</h3><p><a href="https://www.thermofisher.com/in/en/home/industrial/mass-spectrometry/liquid-chromatography-mass-spectrometry-lc-ms/lc-ms-software/multi-omics-data-analysis/compound-discoverer-software.html">Thermo Scientific™ Compound Discoverer™</a> software streamlines compound identification determines real differences between samples and elucidates biological pathways with integrated, powerful software workflows to drive rapid insights from your valuable data.</p><h3>Integration</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*loVT95lJSqN6xqlt.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kM6aIEfTMlTts593l72Bcw.png" /></figure><p>To guide you through the integration workflow, detailed documentation has been prepared and can be <a href="https://docs.elucidata.io/Apps/Metabolomic%20Data/CompoundDiscoverer%20QuantFit.html"><strong>accessed here</strong></a>.</p><p>The node created in Thermo Scientific™ Compound Discoverer™ has 3 integrations:</p><p>1. <a href="http://proteowizard.sourceforge.net/download.html"><strong>MSConvert</strong></a>: It converts .raw data obtained from a mass spectrometer to El-MAVEN accepted formats like .mzML or .mzXML</p><p>2. <a href="https://resources.elucidata.io/elmaven"><strong>El-MAVEN</strong></a>: For high-quality peak integration in a click using a machine learning model. For better peak integration some parameters have been defined in CD.</p><p>3. <a href="https://elucidata.io/platform/"><strong>Polly™ QuantFit</strong></a>: For absolute quantification of the data. This application is hosted on the cloud on Polly™ for processing your data</p><h3>Applications</h3><p>1. <a href="https://resources.elucidata.io/asms#asms2019"><strong>Poster</strong></a>: Determining Pharmacokinetics of Cysteamine with Polly™ QuantFit in collaboration with Shawn Davidson at Princeton University presented at ASMS 2019</p><p>2. <a href="https://resources.elucidata.io/asms#asms2019"><strong>Poster</strong></a>: Rapid automated absolute quantification of metabolites using Polly™ QuantFit to understand tumor nutrient availability in collaboration with Matt Vander Heiden Lab at MIT presented at ASMS 2019</p><p>3. <a href="https://www.youtube.com/watch?v=Ma5Ti3GRayE"><strong>Quant-inar</strong></a>: A webinar on the Polly™ QuantFit Workflow to go from peaks to absolute concentrations within minutes for 100s</p><h3>Free Demo</h3><p>You can now try this integration for free. We offer a free demo on our platform. Drop a message to <a href="/service/mailto:pollydemos@elucidata.io"><strong>pollydemos@elucidata.io </strong></a>today with your requirements and use case. Alternatively, you can directly <a href="https://koi-3qnhzalw1i.marketingautomation.services/net/m?md=EVUKZM86oTvp1OjGgYZ9b0is%2BXysrS3u"><strong>book a suitable time here </strong></a>and talk to us.</p><p><em>Originally published at </em><a href="https://elucidata.io/polly-quantfit-compound-discoverer-integration/"><em>https://elucidata.io</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2133433aabda" width="1" height="1" alt=""><hr><p><a href="https://medium.com/elucidata/the-1-thermo-scientific-compound-discoverer-node-elucidata-2133433aabda">The №1 Thermo Scientific Compound Discoverer Node — Elucidata</a> was originally published in <a href="https://medium.com/elucidata">Elucidata</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Top 10 Latest Single-cell Atlases for Normal Tissues — Elucidata]]></title>
            <link>https://medium.com/elucidata/top-10-latest-single-cell-atlases-for-normal-tissues-elucidata-f567d46de5cf?source=rss-71f407471440------2</link>
            <guid isPermaLink="false">https://medium.com/p/f567d46de5cf</guid>
            <category><![CDATA[single-cell]]></category>
            <category><![CDATA[atlas]]></category>
            <category><![CDATA[data-lake]]></category>
            <category><![CDATA[single-cell-sequencing]]></category>
            <category><![CDATA[top-10]]></category>
            <dc:creator><![CDATA[Elucidata]]></dc:creator>
            <pubDate>Thu, 30 Jul 2020 12:31:00 GMT</pubDate>
            <atom:updated>2020-12-22T10:58:02.298Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Top 10 Latest Single-cell Atlases for Normal Tissues" src="https://cdn-images-1.medium.com/max/1024/1*4PzeZAHUbPgHjqVfWPrYdg.png" /></figure><h3>Top 10 Latest Single-cell Atlases for Normal Tissues — Elucidata</h3><p>By <a href="https://www.linkedin.com/in/richa-mudgal-phd/">Richa Mudgal</a>, <a href="https://www.linkedin.com/in/mdayal/">Manmeet Dayal</a>, <a href="https://www.linkedin.com/in/mrjains/">Vipul Jain</a></p><h3>What is a Single-cell Atlas?</h3><p>A Single-cell atlas is an effort to catalog all different kinds of cells in the human or animal body using the latest techniques in Next-Gen sequencing, microfluidic techniques, nanotechnology, and high-resolution microscopy. Efforts have been made to map every cell that makes up the human body. Such efforts have enabled researchers to characterize individual cells from different organs and tissues and are thought to bring about a change in how we understand, diagnose, and treat human diseases. Such studies provide a roadmap to understanding human health and diseases and form a powerful resource to enable discoveries.</p><p>We have collected the top 10 single cell atlases at hand. Here’s a list comprising the information for 1,902,384 cells from human and mouse tissues. For each atlas, <a href="https://elucidata.io/">Elucidata</a> has compiled its data source, a brief description of the research, together with clickable links to relevant resources. In this article, we give a sneak peek into this ultimate collection.</p><h3>1. Human Cell Landscape</h3><p>In a recent study published earlier this year in Nature, researchers have generated single-cell transcriptomic data profiling over 700,000 cells for 60 human tissue types from adult and fetal tissue. Comparative studies of human and mouse single-cell landscapes helps in identifying the conserved genetic networks components amongst them. The study also reveals that stem and progenitor cells exhibit strong transcriptomic stochasticity, while the differentiated cells are more distinct.</p><p><strong>CELLS</strong>: 702,968 single cells from 67 human tissues &amp; cell culture types<br><strong>PROTOCOLS</strong>: Microwell-seq protocol<br><strong>DATA SOURCE</strong>: GSE134355<br><strong>WEB PORTAL</strong>: <a href="http://bis.zju.edu.cn/HCL/">http://bis.zju.edu.cn/HCL/ </a><br><strong>CODE</strong>: <a href="https://github.com/ggjlab/HCL/">https://github.com/ggjlab/HCL/</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/552/0*_pvUokj9JwJFCyVh.png" /></figure><h3>2. Human Lung Atlas</h3><p>Human Lung Atlas is a comprehensive molecular cell atlas of the adult human lung created using Single-cell RNA sequencing of ∼75,000 cells. The study identifies 58 human lung cell populations including 41 of 45 previously known cell types or subtypes and 14 new ones.</p><p><strong>CELLS</strong>: Cells: ~75,000 human lung and blood cells<br><strong>PROTOCOL</strong>: Droplet- and plate-based single-cell RNA sequencing<br><strong>DATA SOURCE</strong>: Synapse — accession syn21041850<br><strong>WEB PORTAL</strong>: <a href="https://hlca.ds.czbiohub.org/">https://hlca.ds.czbiohub.org/ </a><br><strong>CODE</strong>: <a href="https://github.com/krasnowlab/HLCA">https://github.com/krasnowlab/HLCA</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/485/0*-IysuFRgzqLFkoW_.png" /></figure><h3>8 more Single-cell atlas..</h3><p>Our collection offers 8 more similar atlases ranging from Human Liver Atlas, Mouse Brain Atlas, and more. To download the complete atlas please fill out the form below.</p><p><em>Originally published at </em><a href="https://elucidata.io/top-10-latest-single-cell-atlases-for-normal-tissues/"><em>https://elucidata.io</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f567d46de5cf" width="1" height="1" alt=""><hr><p><a href="https://medium.com/elucidata/top-10-latest-single-cell-atlases-for-normal-tissues-elucidata-f567d46de5cf">Top 10 Latest Single-cell Atlases for Normal Tissues — Elucidata</a> was originally published in <a href="https://medium.com/elucidata">Elucidata</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[When going to office was a thing! — Elucidata]]></title>
            <link>https://medium.com/elucidata/when-going-to-office-was-a-thing-elucidata-5bf0c4244bfa?source=rss-71f407471440------2</link>
            <guid isPermaLink="false">https://medium.com/p/5bf0c4244bfa</guid>
            <category><![CDATA[wfh]]></category>
            <category><![CDATA[company-culture]]></category>
            <category><![CDATA[office-space]]></category>
            <category><![CDATA[office]]></category>
            <category><![CDATA[work-from-home]]></category>
            <dc:creator><![CDATA[Elucidata]]></dc:creator>
            <pubDate>Wed, 29 Jul 2020 12:26:00 GMT</pubDate>
            <atom:updated>2020-07-29T12:26:00.988Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Office space. When going to office was a thing! — Elucidata" src="https://cdn-images-1.medium.com/max/1024/1*8k6EwrrORj6j9unDFkLfig.jpeg" /></figure><h3>When going to office was a thing! — Elucidata</h3><p>We’ve all done it — working remotely a day, a week, every now and then. We all found it comfortable and actually more productive than working in those cubicles. No morning rush, fewer interactions, focused mind, eat as you like, no need to dress-up.</p><p>But four months stuck at home on our own, or with the whole family? We never thought it would be that stressful. Although we are in our comfort zone but away from friends, co-workers. It does affect us. Knowingly or unknowingly office had become our second home.</p><p>Several weeks into this new normal, we find ourselves craving little things like social interaction with our co-workers and being able to saunter to the pantry for a midday coffee break.</p><p>Here are <strong>6 things we miss about working in our office</strong></p><h4>1. Rushing for Shuruate-hafta (our Monday morning company-wide huddle)</h4><figure><img alt="Rushing to Monday morning meetings" src="https://cdn-images-1.medium.com/max/320/0*801GR1vQYXdNZFeN.gif" /></figure><p>Those Monday rushes that gave us gushes. Right from getting on the metro to agreeing to whatever price <em>autowala bhaiyas </em>would cite to reach office on time. Everything was calculated. Now, the Monday meeting is just a click away. We don’t know whether to be happy or sad?</p><h4>2. Lunch chit-chats</h4><figure><img alt="Lunch chit chats. When going to office was a thing! — Elucidata" src="https://cdn-images-1.medium.com/max/1024/0*5-Y06ftqODCR3Ysv.jpg" /></figure><p>When it comes to lunch, you never know what might be on the table<em>. </em>A lunch club can help you get out of the sad-desk-lunch rut, and it comes with a boatload of additional benefits-scientifically-backed. <em>The family that eats together, stays together!</em></p><h4>3. Water-Cooler Gapshap</h4><figure><img alt="water-cooler coversations. When going to office was a thing! — Elucidata" src="https://cdn-images-1.medium.com/max/350/0*AxgNNToo-TPKSvtk.gif" /></figure><p>Taking breaks to refill your water bottle was just a way to chitchat with your colleagues. Whether to rant about a roadblock in a project or just discuss ongoing news. Those breaks not only used to quench your thirst but also refresh your mind. Some of us had become water cooler buddies!</p><h4>4. Those ‘much needed’ Breaks</h4><figure><img alt="When going to office was a thing! — Elucidata" src="https://cdn-images-1.medium.com/max/1024/0*t87WDl2lE4NyEVs7.jpg" /></figure><p>Even if you were stuck at your desk for most of the workday, small adventures like heading out to nearby cafes or parks, or external meetings were used to provide much-needed doses of getaway time. Playing Kho-Kho or football in the park worked like a stress-relieving pill.</p><h4>5. Evening snacks</h4><figure><img alt="Snacks. When going to office was a thing! — Elucidata" src="https://cdn-images-1.medium.com/max/1024/0*c-bwCyT5saAqV9n0.jpg" /></figure><p>Whether your energy levels were taking a dip or you just needed something to munch on while typing away, Whatsapp-ing snacks’ order on our ‘Khan-Pan’ group and visits to the office pantry were a simple joy to look forward to each day.</p><h4>6. Friday party goals</h4><figure><img alt="Friday goals. When going to office was a thing! — Elucidata" src="https://cdn-images-1.medium.com/max/1024/0*ykd0gbcTXBb4maCr.jpg" /></figure><p>Friday used to be the best day of the week, especially after 6 PM. And if you don’t agree, then ask us how we used to feel about it.<em> Ek weekend ki keemat tum kya Jano Ramesh Babu? (You don’t realize the worth of a weekend)</em>.<br>When going out for meal gatherings and bonding with your teammates over some drinks used to be a thing. It was <strong>Friyay</strong>, after all!</p><h4>Bonus Point — Company events leading to get-togethers</h4><figure><img alt="Get-togethers. When going to office was a thing! — Elucidata" src="https://cdn-images-1.medium.com/max/1024/0*6xTpBK1YTHl_dDB5.jpg" /></figure><p>Be it a podcast recording, or an interview, or conducting a webinar, it was always exhilarating to work in-person with our colleagues towards a goal.</p><p>While we have adapted to the new normal, we can’t wait to get back to the <strong>old normal</strong>! Until then even you can stay connected with us via <a href="https://elucidata.io/">Intercom Chat</a>, <a href="https://www.linkedin.com/company/elucidata-corporation">LinkedIn</a>, <a href="https://twitter.com/elucidatacorp">Twitter</a>, <a href="https://www.facebook.com/ElucidataCorporation/">Facebook</a>, or <a href="https://www.instagram.com/elucidata.corp/">Instagram</a>.</p><p><em>Originally published at </em><a href="https://elucidata.io/when-going-to-office-was-a-thing/"><em>https://elucidata.io</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5bf0c4244bfa" width="1" height="1" alt=""><hr><p><a href="https://medium.com/elucidata/when-going-to-office-was-a-thing-elucidata-5bf0c4244bfa">When going to office was a thing! — Elucidata</a> was originally published in <a href="https://medium.com/elucidata">Elucidata</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[7 Reasons why you should use El-MAVEN on Polly — Elucidata]]></title>
            <link>https://medium.com/elucidata/7-reasons-why-you-should-use-el-maven-on-polly-elucidata-c6019d5ee6bf?source=rss-71f407471440------2</link>
            <guid isPermaLink="false">https://medium.com/p/c6019d5ee6bf</guid>
            <category><![CDATA[open-source]]></category>
            <category><![CDATA[maven]]></category>
            <category><![CDATA[mass-spectrometry]]></category>
            <category><![CDATA[metabolomics]]></category>
            <category><![CDATA[data]]></category>
            <dc:creator><![CDATA[Elucidata]]></dc:creator>
            <pubDate>Mon, 27 Jul 2020 08:31:00 GMT</pubDate>
            <atom:updated>2020-07-27T08:31:00.884Z</atom:updated>
            <content:encoded><![CDATA[<h3>7 Reasons why you should use El-MAVEN on Polly — Elucidata</h3><p>by <a href="https://www.linkedin.com/in/jainik-dedhia/">Jainik Dedhia</a></p><p><a href="https://resources.elucidata.io/elmaven">El-MAVEN</a> is an open-source mass spectrometry data processing engine that is optimal for isotopomer labeling and global metabolomic profiling experiments. Currently, it exists as a desktop application that runs on Windows, Linux, and Mac systems. The software can be used to view the mass spectra, align chromatograms, perform peak-feature detection, and alignment for labeled and unlabeled mass spectrometry data. The aim of this software package is to reduce the complexity of metabolomics analysis by using a highly intuitive interface for exploring and validating metabolomics data.</p><figure><img alt="7 Reasons why you should use El-MAVEN on Polly — Elucidata" src="https://cdn-images-1.medium.com/max/1024/0*xZDhejmOTFi7B7t5.png" /></figure><p>Often the size of metabolomic data from an experiment is so large (100 GB or more!) that it becomes almost impossible to process data on a local computer. To overcome this challenge, we have hosted El-MAVEN on a Polly, a cloud-based platform serving all the needs of Biomedical data. Polly can spin up as many numbers of El-MAVEN instances on a variety of cloud machines as needed at a time so that you and your colleagues can simultaneously work on multiple datasets. To know more about Polly check out <a href="https://elucidata.io/platform">this page</a>.</p><h4>1. No Installation required</h4><p>No need to spend time setting up El-MAVEN on a machine and updating to the latest version with every new release. Just log in to Polly from any computer with an internet connection and start using it.</p><h4>2. Choose from a variety of machines</h4><p>The computational power required by El-MAVEN varies with the type and size of data. Polly provides a variety of various machines to choose from, based on your data. Machines with RAM up to 120GB are available.</p><h4>3. Scale for as many users as needed</h4><p>In most cases, labs have one computer with high specifications reserved for such analyses shared between multiple researchers. Polly spins up El-MAVEN instances as per need so that multiple researchers can process their data simultaneously instead of waiting for their turn on a lab computer.</p><h4>4. Ingest data from various sources</h4><p>The instance on Polly provides access to the terminal as well as a browser that can be used to ingest data from a variety of sources from Polly Projects to Cloud Storage solutions.</p><figure><img alt="7 Reasons why you should use El-MAVEN on Polly — Elucidata" src="https://cdn-images-1.medium.com/max/1024/0*rKvbze34jhzqfVHx.png" /></figure><h4>5. One platform for raw data, processed data, and analyses</h4><p>Raw data, processed data, and downstream analyses can be saved on the same platform. Upload raw data from the mass spec machine to Polly and use the El-MAVEN instance to convert data to binary format, pick peaks, and transfer data back to Polly Projects. Use the data for further downstream processing on the platform itself.</p><h4>6. Convenient collaboration</h4><p>Share raw data, processed data, downstream analyses, reports together by simply sharing the Polly project with collaborators.</p><h4>7. Polly support</h4><p>Get access to the Polly Support team for all your queries and help with your analyses.</p><h3>Get a Free Demo</h3><p>El-MAVEN is free to download <a href="https://resources.elucidata.io/elmaven"><strong>here</strong></a>. To give El-MAVEN on Polly a try, <a href="https://koi-3qnhzalw1i.marketingautomation.services/net/m?md=EVUKZM86oTvp1OjGgYZ9b0is%2BXysrS3u"><strong>book a free demo </strong></a>with one of our experts.</p><p><em>Originally published at </em><a href="https://elucidata.io/7-reasons-why-you-should-use-el-maven-on-polly/"><em>https://elucidata.io</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c6019d5ee6bf" width="1" height="1" alt=""><hr><p><a href="https://medium.com/elucidata/7-reasons-why-you-should-use-el-maven-on-polly-elucidata-c6019d5ee6bf">7 Reasons why you should use El-MAVEN on Polly — Elucidata</a> was originally published in <a href="https://medium.com/elucidata">Elucidata</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>