INTRODUCTION

At SVAI, we want to facilitate and support a research community for underserved patients — focused on the undiagnosed and rare disease community. ⭐

More specifically, we’re interested in the expandability of collaborative research and new models of how discovery happens. Our research series started with one patient with a rare neuro-cancer: NF2, then we’ve hosted another successful AI Genomics hackathon focused on p1RCC.
Now, we will host the first hackathon for a patient w/ a heart of gold and undiagnosed metabolic disease: Undiagnosed-1. Sign up here! ✔️

One of my main goals at SVAI was to improve the transparency, interpretability, and effectiveness of the community research done at SVAI. Since I’ve joined, I identified four major bottlenecks in doing the research:‌

1. Understanding the Data
2. Hacking on the Cloud (specifically, Google Cloud Platform)
3. Understanding the biological aspect of the research
4. Understanding the artificial intelligence aspect of the research

Also since then, I created an open, educational platform to address these bottlenecks called Research to the People, v.1.0.

Why this Platform? 🤔

Research to the People hosts resources and knowledge to tackle these major bottlenecks, so that we can then effectively deep dive into illuminating relevant tools and approaches. The hope is then that with the help of your talented mind collaborating with other talented minds in this community, we creatively repurpose and utilize these relevant tools and approaches to tackle any of these three major areas of research (relevant to the patient):

1. patient categorization
2. fundamental biological processes underlying human diseases
3. treating, or developing new treatments.

Addressing the Bottlenecks 🍾

1. Understanding the Data 📊

We dive into the data that allows our research community to yield robust, promising insights into understanding a patient’s clinical case. Depending on the type of disease the patient has, we gather a unique portfolio of data. That data may be genomic/genetics, transcriptomic, proteomic, metabolic, and/or clinical.

Minimally, we try to gather clinical and genomic data. A great next plus would be transcriptome data.

2. Hacking on the Cloud (specifically, Google Cloud Platform) ☁️👩‍💻

In order to access data, and hack onto it, you need to be assigned to a project on Google Cloud Platform (GCP). Once you have a project on GCP, you’ll learn how to use the Storage Engine and the Compute Engine. You’ll mainly be guided through how to store, access your data (in buckets) and how to compute on the data in the cloud (on a VM).

3. Understanding the biological aspect of the research 🧬

Understanding biology leads to majority of groundbreaking insights into biomedicine. The point of this section is to get you familiar with fundamental concepts underlying many powerful approaches to better understand a particular disease or discover relevant insights.

4. Understanding the artificial intelligence aspect of the research 🤖

Machine learning and deep learning approaches have been fundamental to many groundbreaking insights into biomedicine. The point of this section is to get you familiar with fundamental concepts underlying many powerful algorithms — basically, you’re getting some crash courses. This section is by no means comprehensive, but don’t worry, we’ll share a few potentially helpful resources in this section.

Tying it all together 🎁

Finally, what does a good marriage between biology and AI look like? In two sections, we’ll share the a toolbox, the AI-Bio Toolbox, and our most popular application, Cancer Analysis Approaches.

• AI-Bio Toolbox: relevant and very powerful tools for applying AI towards biomedicine. The tools introduced here are biased towards deep learning approaches. You might see some of these packages applied in code-walkthroughs throughout this notebook. (Note: This overview is definitely not comprehensive — a pretty good list of deep learning tools I’ll reference quite a lot exists in this github.)
• Cancer Analysis Approaches: We summarize traditional ML for Cancer, DL for Cancer, and deep-dive into how to (1) Generate and Classify Mutational Signatures, and how to (2) De-noising autoencoders to extract features from breast cancer gene expression data.

Potential Next Steps 🐾

You’ll notice some empty sections — I haven’t had the opportunity to make this as comprehensive as I’d like, but this might be your opportunity to contribute to this open community fascinated with future of computation and the expandability of Biomedical-AI scientific collaboration.

Have fun exploring Research to the People and let me know if you have thoughts for improvement! ✌️

📢P.S. SVAI is hosting its first hackathon for a patient w/ a heart of gold and undiagnosed metabolic disease: Undiagnosed-1. Sign up here! ✔️

#ResearchToThePeopleEducation: An open, educational platform for better biomedical research… was originally published in SVAI on Medium, where people are continuing the conversation by highlighting and responding to this story.

Introduction

Following the first hackathon and its significant momentum for NF2 research and our NF2 patient community partner, Silicon Valley Artificial Intelligence (SVAI) hosted its second annual hackathon: p1RCC Cancer Genomics Research. We welcomed brilliant scientists, engineers, and students from all over North America, and Canada — and were blown away by the talent and results from the weekend.

What does it take to bring over 150 participants together to solve for a rare disease with no known treatment? Let’s dive right in.

Inspiration Behind the Hack

After our successful first hackathon event, NF2, a key hackathon contributor, Bill Paseman, reached out to us: he was diagnosed with stage IIIB papillary renal-cell carcinoma type 1 (p1RCC).

Bill’s diagnosed disease falls under the carcinoma family: Papillary renal-cell carcinoma (pRCC). pRCC accounts for between 15 to 20% of all kidney cancers. It occurs in the cells lining the small tubules in the kidney that filter waste from the blood and make urine. Little is known about the genetic basis of sporadic pRCC, and no effective forms of therapy for advanced disease exist.

Realizing Bill was racing against time, SVAI jumped into deep conversations with Bill and after eight months, SVAI announces its second computational cancer genomics event in SVAI’s Collaborative Research Series, p1RCC, in partnership with RareKidneyCancer.org, Salesforce, Google, NIH, NCBI and more.

The Event Itself

SVAI hosted over 150 researchers, engineers and enthusiasts at Salesforce in San Francisco for an intense weekend of exploration in computational biomedicine. Interdisciplinary teams of computer scientists and biologists collaborated to further understand the disease, for development of potential interventions and to advance the standard of care for p1RCC. In addition to sequencing Bill’s genomic data for this event, SVAI used genomic datasets for p1RCC through the NIH’s Cancer Genome Atlas.

Major Gold-striking: “Nexflix for genes, genomic deep learning, …”

Let’s dive into the deep end of this pool — the scientific advancements that came out of this event. All teams did brilliant work, which can be found here, but a few teams stole the show with some amazing discoveries:

DeeperDrugs —Implementation of rigorous variant filtering and target pruning, including a CRISPR/Cas9 repair design, that pipelines into drug discovery with deep learning, which includes training a DeepChem graph convolutional model, searching for optimal hyperparameters, and applying downstream experimental verification.

AIzheng — Modeled TCGA-RCC tumors as a “time series” across stage (Normal, I, II, III, IV) for each subtype of p1RCC: KIRC, KIRP, KICH. This team evaluated ability of a neural network to discriminate across subtype and stage, constructed stage-specific co-expression networks, and finally identified shared gene interaction communities across each tumor stage.

RecausalNucleotideNetworks — Evaluated the analytical methods being used on BGI-SEQ and to explore the possibility to improve tools for this data type by training and applying genomic deep learning models, including training a BGI-SEQ model for Clairvoyante.

DamTheRiver — Identified of neo-antigens present within patent P1RCC sequence data by machine learning major histocompatibility complex affinity tool. This team implemented a very clean, sophisticated pipeline that ultimately identified 25 peptides with high MHC binding affinity.

ExpressForce — “Nexflix For Genes”: Provided candidate biomarkers for p1RCC via a collaborative filtering using probability matrix factorization after obtaining data from COSMIC, a well-known online cancer catalogue. As an overview, this team created a Sample ID vs Gene matrix table, implemented Naive Bayes algorithm, created entity embeddings of categorical variables, and applied dimensional reduction to find candidate biomarkers for p1RCC.

Overall, these highlighted teams contributed greatly into out understanding of p1RCC, relevant genomics variants and biomarkers of p1RCC, and the accuracy of genomic tools involved in generating and analyzing p1RCC-related data. These teams did outstanding work, and so did all the other teams! I highly recommend checking out the unique discoveries and approaches made by all the teams here.

Gold is good, Diamond is Better

The hackathon was an outstanding success and the teams were all too brilliant, but as movers and shakers in this world, a few teams discovered a drive to dive deeper into the mysteries of p1RCC:

DeeperDrugs, HelloKidney2, and KidneyBean.

In coordination with Bill Paseman and RareKidneyCancer.org, SVAI currently strives to provide more structure, support and visibility to the projects started at our research events. More news to come! Check out SVAI’s research page for the latest updates.

Are You In?

SVAI is a 501(c)(3) non-profit dedicated to building a strong collaborative community committed to accelerating AI for computational biomedicine.

Here’s how you can get involved:

1. Join our community. ❤ Participate in our ongoing and future activities!
2. Donate: Your support helps us sustain and grow our activities.
3. Something else in mind? (*cough* COLLAB?) You’re welcome to contact me at lily@sv.ai

Thanks for reading! Let us know your thoughts.

Cheers,

Lily

Special shoutout to the main event organizers and collaborators: Bill Paseman, Clayton Mellina, Ben Busby, Sean Davis, Pete Kane, Nina Sardesh, Ryan Leung, Hunter Dunbar, Annabelle Tang, and Ripley Jene.

Sprint for the Cure: p1RCC Hackathon was originally published in SVAI on Medium, where people are continuing the conversation by highlighting and responding to this story.