<?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[Kurma Partners - Medium]]></title>
        <description><![CDATA[We are a life sciences-focused VC, investing across biotech, healthtech, techbio, diagnostics and agtech/foodtech. Enjoy the read and please share your thoughts and comments - Medium]]></description>
        <link>https://medium.com/kurma-partners?source=rss----5ae26571354f---4</link>
        <image>
            <url>https://cdn-images-1.medium.com/proxy/1*TGH72Nnw24QL3iV9IOm4VA.png</url>
            <title>Kurma Partners - Medium</title>
            <link>https://medium.com/kurma-partners?source=rss----5ae26571354f---4</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 03:55:57 GMT</lastBuildDate>
        <atom:link href="https://medium.com/feed/kurma-partners" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[From high throughput screening to AI: the quest to transform pharmaceutical R&D]]></title>
            <link>https://medium.com/kurma-partners/from-high-throughput-screening-to-ai-the-quest-to-transform-pharmaceutical-r-d-f5c4c21e9f7d?source=rss----5ae26571354f---4</link>
            <guid isPermaLink="false">https://medium.com/p/f5c4c21e9f7d</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[drug-development]]></category>
            <category><![CDATA[drug-discovery]]></category>
            <category><![CDATA[biotech]]></category>
            <category><![CDATA[biopharma]]></category>
            <dc:creator><![CDATA[Benjamin Belot]]></dc:creator>
            <pubDate>Wed, 03 Jul 2024 08:19:50 GMT</pubDate>
            <atom:updated>2024-07-02T14:18:09.066Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*gSHDKBBknyiFfV6Cdk7cHw.jpeg" /></figure><p>In a data-driven world obsessed with productivity and efficiency, it makes sense for the pharmaceutical industry to consider the use of artificial intelligence, particularly in R&amp;D.</p><p>Having been in the venture capital industry for a couple of decades though, I’ve experienced several innovation waves, bringing the same kind of excitement that AI brings today. Attempting to use this experience to wisely invest with our innate (and certainly not artificial) intelligence, I decided to look back to the early days of my career when genomics, combinatorial chemistry, and high throughput screening were expected to revolutionize new molecule discovery through rationalization.</p><p>Until the 1980s, most drug candidates were small chemistry-based molecules. The discovery effort was then manageable for a pharmaceutical group. Advancements in biology came enriching the therapeutic arsenal with more complex approaches and molecules, such as monoclonal antibodies, gene therapy, cell therapy, and nucleic acids.</p><p>Each of these categories could be further divided into sub-categories. For instance, monoclonal antibodies can be conjugated to toxic molecules or be bispecific. The same applies to gene therapy, which can be permanent, transient, etc.</p><p>Furthermore, production methods are unique to each type of molecule and require significant investments. What was exclusively the domain of chemistry in the 20th century grew to also involve biochemistry or genetic engineering, requiring substantial dedicated investments.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/450/1*7Pi7tggNWqStvXl7WEtNvQ.jpeg" /><figcaption>Reactors used for the production of monoclonal antibodies</figcaption></figure><p>Meanwhile, the threat of blockbuster patent cliffs (i.e. expiration of patent for molecules generating an annual turnover of more than $1 billion) has been looming on the industry.</p><p>With the above in mind, pharmaceutical groups were forced to make decisions to avoid diluting R&amp;D efforts. This led the industry to gradually focus on development and to move further from pure research. A popular option being to acquire innovative companies that discovered molecules of interest, especially those at relatively advanced clinical development stages</p><p>At the end of the last millennium, many emerging biotech companies promised to speed up the discovery of drug candidates, using technologies such as high throughput screening, combinatorial chemistry or gene editing. Three business models coexisted: (i) discovering drug candidates for their own account, then developing them in house or licensing them to the pharmaceutical industry for staggered remuneration, (ii) providing high value drug discovery/development services, (iii) a hybrid model between (i) and (ii).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jXGtctBFU-XzTlkgWa4DrA.jpeg" /><figcaption><em>Robot used for high throughput screening</em></figcaption></figure><p>I preferred the first model back then. My first ever venture was a seed investment in Alantos in 2000, which was eventually acquired by Amgen in 2007 for $300M. Through dynamic combinatorial chemistry invented by Pr Jean-Marie Lehn (Nobel Prize in Chemistry in 1987), Alantos discovered a DDP IV (an enzyme playing a key role in glucose metabolism) inhibitor that was in preclinical development at the time of acquisition. To this day though, I struggle to recall a company of this type transforming into a fully integrated pharmaceutical research and development company (there might be a few of course).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KmEML2fzWevJhTUyzlRmiA.jpeg" /><figcaption>Jean-Marie Lehn, recipient of the Nobel Prize in Chemistry in 1987 together with Donald Cram and Charles Pedersen</figcaption></figure><p>AI-first companies specializing in molecule discovery face a similar choice: either providing services to the industry or developing a proprietary pipeline. The profiles and returns for an investor are very different in each case.</p><p>The service model is less likely to generate high returns but doesn’t require extensive capital resources. On the other hand, developing a pipeline of proprietary products, even at an early stage, while perfecting the tech is extremely capital intensive. And the chances of discovering a blockbuster are slim, hopefully less so with the use of AI.</p><p>AI-driven drug discovery companies out there seem to be leaning towards the second option, taking the challenge! Not a week goes by without the release of a new (multimodal) LLM for biology, from AlphaFold 3 (Isomorphic Lab) to ESM3 (EvolutionaryScale) to Tx-LLM (Google Deepmind). But the sheer complexity of biology is unmatched and the data available to train model is scarce, as opposed to what can be done in natural language.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GxeeZ2eZilCGR65fnqtWkw.jpeg" /><figcaption>AlphaFold 3 predicts the structure of proteins, DNA, RNA, ligands…and how they interact</figcaption></figure><p>Keeping this complexity in mind, it is difficult to believe that artificial intelligence can address all the challenges of drug discovery &amp; development at once. Perhaps the question we should ask ourselves is where in the R&amp;D pipeline (target identification, hit-to-lead, toxicology…) do we expect AI to have the biggest impact. And focus on extracting the most value possible there.</p><p><em>Written by Philippe Peltier with contribution from Benjamin Belot</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f5c4c21e9f7d" width="1" height="1" alt=""><hr><p><a href="https://medium.com/kurma-partners/from-high-throughput-screening-to-ai-the-quest-to-transform-pharmaceutical-r-d-f5c4c21e9f7d">From high throughput screening to AI: the quest to transform pharmaceutical R&amp;D</a> was originally published in <a href="https://medium.com/kurma-partners">Kurma Partners</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Exploring “TechBio” —  what’s behind the hype?]]></title>
            <link>https://medium.com/kurma-partners/exploring-techbio-whats-behind-the-hype-be5edd4eb3e0?source=rss----5ae26571354f---4</link>
            <guid isPermaLink="false">https://medium.com/p/be5edd4eb3e0</guid>
            <dc:creator><![CDATA[Laura Achach]]></dc:creator>
            <pubDate>Tue, 23 Jan 2024 09:01:30 GMT</pubDate>
            <atom:updated>2024-01-23T09:01:30.325Z</atom:updated>
            <content:encoded><![CDATA[<h3>Exploring “TechBio” — what’s behind the hype?</h3><p>In recent months, ‘Generative AI’ and ‘TechBio’ have become hot topics in healthcare innovation circles (see <a href="https://medium.com/kurma-partners/how-foundation-models-may-finally-enable-ai-to-transform-radiology-9076b1850caf">link </a>for Benjamin Belot’s article on the transformative effects of Generative AI in radiology).</p><p>In contrast to “Biotech”, “TechBio” refers to applications of the tech revolution to biology. Biotech and Tech VCs are ramping up their investments in the field to address this surge of innovation. But what is it all about, and why is it becoming hot now? This article seeks to define TechBio, discuss how it may defer from Biotech, and uncover the reasons behind the recent surge of interest.</p><h3><strong>General Definition and Examples</strong></h3><p>Let’s start by defining TechBio. This emerging sector combines engineering and computer principles to address biological challenges, paving the way for innovative treatments and diagnostic solutions.</p><p>For a clearer understanding of what falls within this sector, see below for a breakdown of the TechBio space with examples of companies:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QgNj0196juri55NWn3Tybw.png" /></figure><h3><strong>Decoding TechBio vs Biotech</strong></h3><p>Biotech focuses on applying engineering (like tissue, and molecular engineering) to develop biological assets, while TechBio companies create tools that have an impact on biology to support and accelerate the development of new drugs or diagnostics.</p><p>However, it’s primarily the model of TechBio startups that sets them apart from Biotechs, based on the following three aspects:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/937/1*zqogJPU2RYddvkQ3sOmZcA.png" /></figure><p><strong>Management team</strong></p><p>From a management perspective, the typology of entrepreneurs at the helm often differs.</p><p>In Biotech, most promising technologies are coming from top-tier academic institutions and rely on decades of discoveries on cellular pathways or new mechanisms of action for example. These innovations are converted into products within biotech startups. Even though expert scientists are behind innovations, they are often joined by experienced CEO, most of the time seasoned entrepreneur, to lead the company’s development. Some biotech funds, such as Kurma Partners, have even opted for the start-up studio model, scouting out promising scientific discoveries from academic labs and supporting formation of management teams. This is what they did, for example, with EmergenceTx, a company recently acquired by Eli Lilly (<a href="https://www.biopharmadive.com/news/lilly-acquire-emergence-adc-cancer-drugs/654326/">link</a>).</p><p>TechBio companies, on the other hand, are often founded and run (at least in the early stages of development) by young founders with diverse backgrounds in technology, engineering, and biology. Many of these creative founders are first-time entrepreneurs who have quickly identified a market opportunity and carried out rapid iterations to develop a minimum viable product (MVP) and bring innovations to the world.</p><p>A prime example is Kurma Partners portfolio company DNA Script. It was founded by a team of young engineers and scientists with the ambition to print DNA on a bench. To reach this goal, the company assembled the best of molecular biology, hardware engineering, and software development.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/612/0*Bf2B-LeT9OLSFZ0k.jpg" /><figcaption>Co-founders of DNA Script (Credit: DNA Script)</figcaption></figure><p><strong>Platform vs asset-centric Model</strong></p><p>Another key distinction between TechBio and Biotech lies in the platform-centric approach of some TechBio companies.</p><p>Biotech companies are typically asset-based companies that have a specific initial intellectual property on a new form of therapy/compound.</p><p>AI-enabled drug discovery (AIDD) solutions, on the other hand, are often platform companies that focus on a technological approach or system that should enable them to produce several new drugs across various indications. These platform-centric companies are not only defined by their asset pipeline; instead, they are characterized by their robust discovery engines.</p><p>However, while AIDDs begin by focusing on platforms, they may evolve into a hybrid of the two models and develop their proprietary asset pipeline. For instance, some AIDDs start with a platform model, generate early revenue by aiding crucial R&amp;D processes for other companies, and simultaneously accumulate the necessary data and expertise to create proprietary assets (co-owned or fully owned). A notable example is <a href="https://insilico.com/">Insilico Medicine</a>, which has evolved into a clinical-stage, end-to-end AI drug discovery company and has built a pipeline of drug candidates.</p><p><strong>Capital requirements</strong></p><p>While the path to market for biotech is capital intensive due to the various stages of technological, clinical, and regulatory development, TechBio companies may prove to be more capital efficient. They can leverage technology to de-risk and bring products/services to market faster with funding cycles/ rounds similar to software companies. This similarity with the software model is probably what caught the eye of some of the more generalist Tech funds out there.</p><p>A notable exception could be AIDDs, which need capital for both technology and asset development early on. However, the assumption is that the next generations of hits/candidates will be cheaper to churn out.</p><h3><strong>Why now?</strong></h3><p>The TechBio landscape is on the upswing, driven by a combination of technological advancements and shifts in industry dynamics. We’ve witnessed the following key transformations.</p><p><strong>Shift in mindset within the industry: design-first approach. </strong>Traditionally the drug discovery process involved searching for a specific biological target and a drug that affects it, screening millions of potential drugs. Now, companies can simulate structures and dynamics, virtually screen libraries of compounds, and also lean towards <strong>intentional drug design</strong>. In novel drug design, they can start with desired properties and use computational techniques to find or design the right drugs.</p><p><strong>Increased availability of data:</strong></p><p>The <strong><em>cost of genomic data has significantly reduced</em></strong>, making genomic knowledge more accessible and widely used. Companies like <a href="https://www.illumina.com/">Illumina</a>, which played a big role in modern genomics, along with others like <a href="https://www.10xgenomics.com/">10X Genomics</a> and <a href="https://nanoporetech.com/">Oxford Nanopore Technologies</a>, are making it easier to understand complex genetic sequences by developing new methods, including long-read technologies.</p><p>It is also now possible to combine genomic data with other omics to better understand the relationship between systems and hence the underlying mechanisms of diseases.</p><p>More and more companies are creating virtuous feedback loops by using available data to generate a model and testing the model’s predictions in wet lab conditions…thus generating additional proprietary data that can further improve the model’s accuracy</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/720/1*RurdU9ju9nIbwKYp1bUSzQ.png" /><figcaption>Absci’s Integrated Drug Creation™ platform (Credit: Absci)</figcaption></figure><p><strong>Artificial intelligence (AI) and machine learning (ML) advancements </strong>play a crucial role in analyzing biological data. Recent breakthroughs, in large language models (LLMs) and generative AI (GenAI), enhance our understanding of biological complexity.</p><p>Some examples here:</p><p>· <a href="https://www.cradle.bio/">Cradle.bio</a> uses generative AI to design new proteins by capturing the link between protein sequence and structure.</p><p>· We are also seeing a trend towards the development of multimodal foundation models i.e. able to ingest and analyze multiple data sources. For example, <a href="https://www.forbes.com/sites/kenrickcai/2023/08/25/evolutionaryscale-ai-biotech-startup-meta-researchers-funding/?sh=773e6aa3140c">EvolutionaryScale</a> wants to go beyond predicting protein structure, as <a href="https://alphafold.ebi.ac.uk/">Alphafold</a> does, by developing a multimodal foundation model integrating other biological data from DNA sequences, gene expression, and epigenetic states.</p><p>· In clinical trials, we are also seeing the impact of these technological advances, with companies such as <a href="https://www.novadiscovery.com/">Nova Discovery</a>. Through in silico modeling, they can generate insights enabling smaller and more targeted studies. This allows for a decrease in the required number of patients within control groups or the creation of complete synthetic control arms. Recently, they’ve also successfully predicted AstraZeneca’s phase 3 clinical trial outcomes in advance of their announcement with precision and foresight. (<a href="https://www.novadiscovery.com/novadiscovery-announces-success-of-first-of-its-kind-clinical-trial-simulation-to-accurately-predict-findings-of-phase-iii-clinical-study/">see link here</a>)</p><p><strong>Automatization: </strong>new technologies are making it possible to automate biological processes taking place in the lab. <a href="https://automata.tech/">Automata</a>, for instance, is a good example of innovation combining software and hardware to automate scientific laboratories, saving scientists time and effort.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*Nkz4_IzsotWZ5kWZAoreSw.png" /><figcaption>LINQ: The complete lab workflow automation platform (Credit : <a href="https://automata.tech/">Automata</a>)</figcaption></figure><p><strong>Recent regulatory changes encourage the use of more effective human-based models to test new drugs</strong>. The FDA Modernization Act of 2022 no longer requires animal trials, aiming to address ethical concerns as well as the lack of predictive powers of certain animal models. Instead of using animals, companies can now opt for organoids (patient-derived mini organs) in pre-clinical testing. <a href="https://www.orakl-oncology.com/">Orakl</a>, a spinoff of the renowned cancer center Gustave Roussy Institute, is building one of the largest biobanks of tumor avatars from fresh patient biopsies obtained through collaboration with the institute. Avatars can be used to discover new targets, increase drug throughput compared to animal models and screen drug efficacy early to mitigate clinical trial risks for Pharma and Biotech. Their approach merges organoid biology, high-quality clinical data, and AI that provide wet-lab and dry-lab insights that capture the full complexity of cancer.</p><h3><strong>Conclusion</strong></h3><p>Tech is having an ever-increasing impact on bio, and a growing number of companies are attracting the attention of investors. Unlike Biotech, TechBio startups have models similar to software companies, founder-led and platform-focused, they are often able to generate revenue quickly. This is attracting interest from generalists and technology investors interested in life sciences.</p><p>We’ve been in this game since 2015, and we’ve seen an acceleration of innovation driven by a cultural shift in drug development approaches. The focus is on data-driven rather than hypothesis-driven methodologies, propelled by recent advances in AI and ML applied to biology. Opportunities for automation and scalability of tedious processes are generating valuable data, complemented by regulatory support for emerging technologies.</p><p>Our Healthtech team is more than ever interested in meeting and supporting visionary entrepreneurs developing impactful technologies in the space. Having a biotech team is also a big plus as we also benefit from their clear understanding of the nuanced requirements of the biotech and pharma industries. Feel free to reach out!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=be5edd4eb3e0" width="1" height="1" alt=""><hr><p><a href="https://medium.com/kurma-partners/exploring-techbio-whats-behind-the-hype-be5edd4eb3e0">Exploring “TechBio” —  what’s behind the hype?</a> was originally published in <a href="https://medium.com/kurma-partners">Kurma Partners</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How foundation models may finally enable AI to transform radiology]]></title>
            <link>https://medium.com/kurma-partners/how-foundation-models-may-finally-enable-ai-to-transform-radiology-9076b1850caf?source=rss----5ae26571354f---4</link>
            <guid isPermaLink="false">https://medium.com/p/9076b1850caf</guid>
            <category><![CDATA[generative-ai-solution]]></category>
            <category><![CDATA[multimodal-ai]]></category>
            <category><![CDATA[radiology]]></category>
            <category><![CDATA[healthtech]]></category>
            <dc:creator><![CDATA[Benjamin Belot]]></dc:creator>
            <pubDate>Thu, 04 Jan 2024 17:47:46 GMT</pubDate>
            <atom:updated>2023-10-30T14:45:06.148Z</atom:updated>
            <content:encoded><![CDATA[<p>Back in 2016, famous computer scientist, AI expert &amp; Turing Award recipient Geoffrey Hinton was urging the world to “stop training radiologists”. As algorithms were becoming more and more accurate, it seemed like AI could solve the well documented global shortage of radiologists. Fast forward a few years later and it’s fair to say that despite the hopes and the hype, AI has had a somehow limited impact on the practice of radiology.</p><p>With the advent of large language models (LLMs), we could be ushering into a new era for AI in healthcare. Early use cases are already being identified, especially on medical documentation and back-office workflows of care (medical coding, revenue cycle management…). These use cases have an exciting potential to ease the lives of healthcare providers, dedicating their time and energy to helping patients but growing more and more fatigued.</p><p>AI solutions improving the delivery of care itself are admittedly more challenging to develop, but they might be even more impactful. In this article, we will focus on radiology, looking into the main challenges that hindered AI adoption in the first place and how advances in models’ architecture and training methods may significantly move the needle.</p><p><strong>AI in Radiology: why has adoption been limited thus far?</strong></p><p>As of January 2023, out of the 520 AI algorithms approved by the FDA, 396 were pertaining to radiology, making it a key area of focus for AI companies.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/638/1*bsTzVQyRR_34joVvJXVyfg.png" /><figcaption>FDA-approved AI-based medical devices recap (by the Medical Futurist)</figcaption></figure><p>AI promised to revolutionize medical imaging, offering the potential for faster, more accurate, and cost-effective disease detection. Technologically, it achieved significant progress through deep learning, with some algorithms surpassing clinical experts on accuracy in specific tasks (<a href="https://www.nature.com/articles/s41551-022-00936-9">example</a>)</p><p>…so what happened?</p><p>Below we listed some of the key hurdles we have identified.</p><p>Let’s first dig into what may be called the <em>tunnel vision trap</em>. Up to now, models were typically trained through supervised learning, using large amounts of (well) labelled data to learn to recognize patterns and achieve classification tasks. Such data is particularly hard to come by in healthcare: it is both costly and time-consuming to have new data labeled by experts, themselves in short supply. Furthermore, this task-specific approach created inflexible models that are limited to predefined tasks and cannot adapt to other tasks without costly retraining.</p><p>The result is a collection of algorithms detecting a single ailment on a single imaging modality, making it impractical for radiologists to include them in their workflow. Indeed, radiologists must be able to identify dozens of abnormalities in their daily practice and they may be looking for more than one on a given image.</p><p>Furthermore, radiologists don’t reach conclusions on images alone, also incorporating lab test reports, patient history…in their analysis. Traditional AI models struggled to handle this multifaceted information, failing to take stock of all the complexities of the job.</p><p>Another shortfall of supervision-based algorithms is their limited generalization potential. Many early AI models were designed to perform well on specific datasets or tasks but struggled to generalize to new or diverse cases, whereas radiology demands a high level of generalization given the wide variety of patient conditions, workflows and imaging devices in service.</p><p>Challenges related to interpretability and trust should also be mentioned. Radiologists require transparency in AI decision-making. The “black box” nature of deep learning AI models has hindered their acceptance, as healthcare professionals often need to understand the rationale behind AI-generated insights.</p><p>All of the above contributed to cast a shadow of doubt on the real benefits of AI for radiology, even prompting some practitioners to declare that it was adding to their already heavy workload!</p><p><strong>LLMs: how can they move the needle?</strong></p><p>Over the past few years, significant advances have beenmade in machine learning, particularly in model architecture and learning methods. First described in a <a href="https://arxiv.org/abs/1706.03762">Google paper</a> from 2017, transformers are a new generation of neural networks capable of identifying subtle patterns within a dataset, notably between elements that may be sequentially far from each other (e.g. words in a sentence).</p><p>The use of transformers paved the way to train models on large amounts of unlabeled data using self-supervised learning (SSL). As a result, models learn a richer representation of the environment and can then be<em> </em>adapted or fine-tuned to a wide range of downstream tasks.</p><p>In other words, the so-called “foundation model” can then be trained for specific tasks in a much more agile way requiring lesser amounts of well-labelled data — just like ChatGPT can perform summaries, translate text, generate new content…Down the line, the opportunity is for a foundation model to replace many task-specific models.</p><p>Some early use cases of generative AI in healthcare are presented below. As you can see, they mostly address documentation and back-office workflows of healthcare. It’s also noticeable that the competition is already becoming fierce.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/945/1*ZdHu_AQSSXYK72ahZd3hWw.png" /><figcaption>Mapping of front &amp; back office generative AI solutions in Healthcare (by Sequoia Capital)</figcaption></figure><p>In terms of performance, it’s interesting to note that SSL models have outperformed their supervised counterparts on many tasks (<a href="https://www.nature.com/articles/s41586-023-06555-x">example</a>)<em>, </em>despite being fine-tuned with smaller amounts of data.</p><p>Transformers also tend to fare much better when exposed to new data from different domains. As they excel in learning patterns and relationships from extensive and varied datasets, they can be pre-trained on vast medical databases, ensuring robust generalization across different patient demographics and imaging technologies. A robustness which is essential for real-world clinical applications.</p><p>Finally, on the topic of transparency and explainability: LLMs and transformers can incorporate natural language processing (NLP) to analyze and generate textual explanations for their predictions. This feature enhances interpretability, allowing radiologists and healthcare professionals to understand and trust AI-generated recommendations.</p><p>Of course, we must still remain cautious: biases, data privacy or hallucinations (i.e. the tendency of these models to over confidently generate false information) are some of the serious challenges that will need to get dealt with. But the potential to address some of the shortcomings of previous AI models is definitely there.</p><p><strong>Multimodal AI: the next frontier in radiology</strong></p><p>While the first LLMs typically focused on text, the latest generations are multimodal i.e. equipped with “extra sensory skills” to also process images, audio, videos… thus making them closer to human intelligence.</p><p>This is particularly interesting in the context of medicine: an inherently multimodal discipline<em>.</em> When providing care, HCPs typically analyze data from a variety of modalities, such as medical images, clinical notes, blood tests &amp; more. All of these data are combined in clinical decision making: for example a brain lesion may be difficult to characterize without any context and based on imaging features alone. If we want to bring true AI assistants (i.e. copilots) for doctors, they also need to be multimodal.</p><p>As the mighty Eric Topol put it « <em>the big shift ahead is the ability to transcend narrow, unimodal tasks, confined to images, and broaden machine capabilities to include text and speech, encompassing all input modes, setting the foundation for multimodal AI</em> » . Hence the recent interest for models such as <a href="https://arxiv.org/abs/2307.14334">Med-PaLM M</a>, <a href="https://arxiv.org/pdf/2307.15189.pdf">Med Flamingo</a>, <a href="https://www.nature.com/articles/s41586-023-06555-x">RETfound</a> or <a href="https://arxiv.org/abs/2310.09909#:~:text=Our%20observation%20shows%20that%2C%20while,diagnosis%20and%20generating%20comprehensive%20reports.">GPT 4V(ision)</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/945/1*uY1_IhJ33KZpgb3ZXDFlXA.png" /><figcaption>Med-Palm M: multimodal medical LLM for healthcare produced by Google</figcaption></figure><p>Businesswise there is also an opportunity here, in a context where everybody is wondering who will extract value from the generative AI value chain (besides Nvidia that is…) VS be gobbled up by big tech or AI incumbents (this category is so underrated!).</p><p>Large models from big tech players are generalist by design and hence have not been exposed to tons of healthcare specific data. Unsurprisingly they tend to perform less well than the specialized models mentioned above. Multimodal AI opens up the opportunity to create more data moats as well as vertical SaaS-style AI products that will better meet the needs and fit within the workflows of healthcare users.</p><p>At Kurma Partners, we believe in the potential for a second wave of “AI for radiology” companies, leveraging the latest technology developments to completely revamp the approach, from model training to workflow integration.</p><p>Rather than incremental steps, the approach might entail rebuilding an AI-ready multimodal viewer from the ground up, that can be interacted with through prompts (just like ChatGPT) and work across imaging modalities (MRI, CT…)… and more (clinical notes, biology test reports…). Conceptually, this “ground up” approach is similar to what companies developing AI for drug discovery are trying to do: turning the table from an approach based on screening millions of compounds to see which one sticks to a pro-active design approach.</p><p>There are foundational (pun intended) generative AI-native companies to be built in healthcare and we are seeing more and more emerge from stealth, one of them being our own portfolio company <a href="https://www.raidium.eu/">Raidium</a>. Make sure to check it out if the thesis developed in this article resonates.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/945/1*M8N18iT7SJ9UtV9O0qQ5Fw.png" /><figcaption>A view from the Nvidia headquarters in Santa Clara</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9076b1850caf" width="1" height="1" alt=""><hr><p><a href="https://medium.com/kurma-partners/how-foundation-models-may-finally-enable-ai-to-transform-radiology-9076b1850caf">How foundation models may finally enable AI to transform radiology</a> was originally published in <a href="https://medium.com/kurma-partners">Kurma Partners</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A digital health investor’s perspective: digital therapeutics (DTx)]]></title>
            <link>https://medium.com/kurma-partners/a-digital-health-investors-perspective-digital-therapeutics-dtx-d46062b414f2?source=rss----5ae26571354f---4</link>
            <guid isPermaLink="false">https://medium.com/p/d46062b414f2</guid>
            <category><![CDATA[digital-health]]></category>
            <category><![CDATA[early-stage-investing]]></category>
            <category><![CDATA[digital-therapeutics]]></category>
            <category><![CDATA[venture-capital]]></category>
            <dc:creator><![CDATA[Benjamin Belot]]></dc:creator>
            <pubDate>Mon, 22 Mar 2021 16:09:51 GMT</pubDate>
            <atom:updated>2021-03-23T15:52:19.746Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Kicking off this new series of articles, in which we will be sharing views on several digital health-related topics, from our perspective as early-stage investors</em></p><p>Digital Therapeutics (DTx) have been in the spot light for some time now and the picture is starting to get clearer. Multiple start-up/pharma partnerships, FDA approvals and (finally) substantial M&amp;A action (i.e. Teladoc acquiring Livongo for US$18.5 bn in October 2020) are all signs pointing towards the maturation of the sector. Even regulators &amp; payers are working on ways to better assess and support this new class of treatment: the best example being Germany, which enacted the Digital Care Act (DVG) in 2020 to provide early reimbursement to digital therapies. In this article we will share our perspective on some of the opportunities and challenges of DTx.</p><figure><img alt="Digital Therapeutics within the greater ensemble of digital healthfrom CB Insights" src="https://cdn-images-1.medium.com/max/599/1*GzrC5PByGIZB4NB_fZWVMg.png" /><figcaption>The DTx space within Digital Health from CB Insights</figcaption></figure><p><strong>First thing first: what is a digital therapeutic?</strong></p><p>As the name suggests, digital therapeutics are software or apps intending to provide therapeutic benefits, as a standalone or in combination with another modality (drug, physical intervention…). But to earn the DTx label, these benefits must be backed by clinical evidence. As per the Digital Therapeutics Alliance’s definition: <em>Digital therapeutics (DTx) deliver evidence-based therapeutic interventions that are driven by high quality software programs to prevent, manage, or treat a medical disorder or disease. They are used independently or in concert with medications, devices, or other therapies to optimize patient care and health outcomes.</em></p><p>This is fully aligned with our investment philosophy: if you claim to be either a therapeutic or a diagnostic product, you should be prepared to demonstrate your value with rigorous science. As a side note, we do believe that digital health as a whole has suffered from a general lack of evidence, with a multitude of apps and software making bold claims with no semblance of clinical data (i.e. the snake oil syndrome).</p><p><strong>Why so much interest in digital therapeutics?</strong></p><p>By virtue of being software products, DTx provide a scalable form of therapy with less or no side effects compared to pharmacological drugs. In chronic pain for example, the side-effects and risks associated with opioids have been very documented. Governments, healthcare systems and patients are actively looking for safer yet effective alternatives. The solutions developed by our portfolio company Lucine (<a href="https://lucine.fr/">https://lucine.fr/</a>) is providing just that, by using a digital therapy to directly affect the mechanisms of pain known as nociception.</p><p>Overall, we see a lot of potential for DTx in disease areas where standards of treatment leave significant needs unmet and where a mental/behavioral component can be (at least partially) addressed through habit formation.</p><p>Learning disabilities such as dyslexia interfere with basic children’s learning skills that can harm further education and consequently future in adulthood. In addition, irrespective of the quality of care by dedicated therapists, the frequency of care is important. This is one of the reasons why we invested in BMotion Technologies, a DTx company providing a rehabilitation product called Mila (<a href="https://www.mila-learn.com/">https://www.mila-learn.com</a>), using music &amp; games to help children with learning disabilities. These games are engaging and fun to play but that’s not all: they trigger well-known mechanisms in the brain to address cognitive challenges.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/624/1*ibaXwaTEWuTj_7hiokAx1A.png" /><figcaption>BMotion Technologies’ Mila-Learn rehabilitation DTx</figcaption></figure><p>Another benefit of digital therapeutics is the ability toleverage rapid feedback collected from users. Indeed, by monitoring how patients interact with the product, individualizing the intervention and identifying (and solving for) non-compliance issues becomes possible.</p><p><strong>What are (some of) the challenges that lie ahead for DTx companies?</strong></p><p>We talked about the importance of generating clinical evidence, but from a product perspective digital therapeutics must also be engaging enough to foster sustained use. In other words, DTx products must be appealing and offer a satisfying patient experience in order to get compliance.</p><p>With that in mind, gamifications approaches used by Akili Interactive (the first company to develop a therapeutic video game and getting a FDA approval for it) or BMotion Technologies are interesting because they are more likely to keep the user coming. Designing a compelling digital product for a healthcare-related use is not trivial though and itrequires the right skill set. Expect to hire some talented game designers if you intend to create a videogame-like experience.</p><p>Go-to-market is another interesting challenge and one that no DTx player has completely cracked. By-and-large, we can identify 3 go-to-market approaches to DTx:</p><p><strong>Prescription DTx</strong> are digital drugs i.e. drugs which happen to be of a non-pharmacological nature. As such it makes sense that they be prescribed by doctors.</p><p>Pioneers such as Akili Interactive (<a href="https://www.akiliinteractive.com/">https://www.akiliinteractive.com/</a>) or Pear Therapeutics (<a href="https://peartherapeutics.com/">https://peartherapeutics.com/</a>) have chosen this path, going through the most rigorous clinical evidence-generation process and eventually registering their product(s) with the FDA, hence getting a universal stamp of approval for medical relevance. This stamp is a high reward but the road is long, difficult and costly for a start-up that doesn’t have the war chest of a big pharma and is experimenting with a completely new modality.</p><p>Furthermore, regulatory approval is far from being the end of the road: getting doctors comfortable enough with your product to prescribe it is a crucial step (as is securing reimbursement). Hiring a sales force is expensive, hence the option to partner with a pharma company may seem enticing. But pharma companies are good at… selling pharmaceutical products! Selling digital products is a whole different ball game and one that can entail a steep learning curve. The example of Pear Therapeutics and Novartis/Sandoz aborted deal is still in a lot of memories as a wasted opportunity. That being said there has been a series of high profile partnerships in the last few months (Chiesi &amp; Kaia Health, Boehringer Ingelheim &amp; Click Therapeutics…) and it will be interesting to see how they incorporate learnings from the past to create a landmark for success. Not putting all eggs in the same basket (i.e. a portfolio management approach) may be the best way and it’s exactly what Kaia Health (<a href="https://www.kaiahealth.com/">https://www.kaiahealth.com/</a>) is doing. They entered into a strategic commercialization partnership with Chiesi for their COPD product, while retaining the rights of the core chronic back pain/MSK product.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/444/1*Dz6ososCuH_a5KBPZ4h-JQ.jpeg" /><figcaption>Akili Interactive’s therapeutic video game to treat ADHD</figcaption></figure><p><strong>Payers-first DTx</strong> are aiming to leverage healthcare payers rather than doctors in order to reach patients. The approach is more common in the US, where the payer landscape is more fragmented and where payers hold strict formularies of what they cover, thus having more influence on treatment options for their members.</p><p>This is the approach that Livongo (<a href="https://www.livongo.com/">https://www.livongo.com/</a>) favored: by catering to end points that are meaningful for payers (with regards to glucose and HbA1c control in particular), they were able to get contracts signed and revenue coming in, without relying on doctors prescriptions. Although the approach may seem easier, it definitely comes with some downsides. Despite formularies, payers are typically not the most efficient organizations to handle user acquisition. Just because they endorse a treatment doesn’t mean that their members will rush to adopt it, unlike the recommendation of a doctor which typically carries more weight.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/392/1*jEp-webMSgKZ36AL0xtJ6A.png" /><figcaption>Livongo’s suite of products</figcaption></figure><p>Also, with more and more companies seemingly choosing the payers-first path to get to the market faster (i.e. avoid lengthy regulatory processes and/or having to knock on the doctor’s door), there is a risk that payers end up flooded with a multitude of solutions and it’s highly unlikely they will cover all of them.</p><p>Going <strong>direct-to-patient </strong>is another option to reach the market fast. It’s not the easiest approach in Europe though, as citizens are used to so-called “free healthcare”. The other challenge is the hundreds of thousands of direct-to-consumer healthcare apps out there competing for attention and not always resulting in a positive experience. But it can work if you have what it takes to stand out from the crowd and if you are addressing a chronic ailment with a strong impact on quality of life. Take chronic back pain for example: after multiple sessions with a physiotherapist, painkillers and possibly the prospect of a very invasive yet uncertain surgery, no wonder people are willing to pay out of pocket for Kaia Health’s product, which combines cognitive behavior therapy with tailored exercise programs and even remote sessions with a physiotherapist if needed.</p><p>We at Kurma Partners tend to have a sweet spot for prescription DTx, but we recognize that the approaches listed above are not mutually exclusive and the best way is probably a combination of at least two of them. Taking yet again the example of BMotion Technologies, the company has the ambition to demonstrate that its product is prescription-grade, but meanwhile many parents struggling with their child’s dyslexia are taking matters into their own hands and thus looking for treatments now, hence the significant B2C traction that the company has been experiencing.</p><p>Many more challenges remain to be addressed in order to cement DTx’ position as a credible therapeutic option (quality management, distribution to patient, data management…) but the field has been fascinating to follow over the last few years. A lot has happened, from the inception of the concept itself to the first FDA approvals to a first high profile transaction with the acquisition of Livongo by Teladoc. We are convinced that this modality is addressing a need for patients, HCPs and payers alike and is thus here to stay.</p><p>If you are an early-stage DTx company with the ambition to do something big and meaningful, please give us a shout → <a href="http://contact@kurmapartners.com/">contact@kurmapartners.com</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d46062b414f2" width="1" height="1" alt=""><hr><p><a href="https://medium.com/kurma-partners/a-digital-health-investors-perspective-digital-therapeutics-dtx-d46062b414f2">A digital health investor’s perspective: digital therapeutics (DTx)</a> was originally published in <a href="https://medium.com/kurma-partners">Kurma Partners</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>