

Discover more from Where Tech Meets Bio
Why Your Main AI Strategy Should Be Data Connectivity
ALSO: Weekly Highlights; Pick of the Week; Digging Gems of Biopharma
Hi! I am Andrii Buvailo, and this is my weekly newsletter, ‘Where Tech Meets Bio,’ where I talk about technologies, breakthroughs, and great companies moving the biopharma industry forward.
If you've received it, then you either subscribed or someone forwarded it to you. If the latter is the case, subscribe by pressing this button:
Now, let’s get to this week’s topics!
I came across this LinkedIn post by Tony Seale, Knowledge Graph Engineer at UBS, and it conveys such an important idea for all of us who are in the business of integrating predictive and generative artificial intelligence (AI) in our organizations! Here is the entire text with the original graphics (with permission):
In the rapidly evolving landscape of AI, the ability to remain distinct and competitive hinges on an unexpected factor: how interconnected your data is.
Think of your organisation as a living entity—its survival depends on a well-defined information boundary, which functions like a semi-permeable membrane. Karl Friston's Free Energy Principle (FEP) models this boundary as a Markov Blanket and says that to sustain itself, a system must minimise its free energy. Free energy is essentially a measure of surprise or uncertainty, and minimising it equates to maintaining a state of low internal entropy. A system achieves this by forming accurate predictions about the external environment and updating its internal states accordingly, thus allowing for a dynamic yet stable interaction with its surroundings. All of this is only possible because the Markov Blanket delineates a boundary between internal and external systems.
The versatility of the FEP is remarkable, spanning various scales. Chris Fields' application to quantum mechanics, suggests that an entangled quantum system maintains a distinct information boundary, that is only loosely coupled to its environment. An entangled quantum system is a distinct system. The implication is potent: the more interconnected the system's components, the stronger the boundary and thus the more pronounced its individual identity.
⚡The degree of your internal connectivity equates to how distinct you are⚡
Upon scaling this concept to an organisational level, we face a sobering reality. Often, our data is siloed and disconnected, lacking any unifying structure, common vocabulary, or shared semantics. Disconnected systems signal weak boundaries. The lower the data connectivity, the less distinct you are as a whole. Your organisation will unintentionally leak subtle information across its boundary. This did not matter before, but advanced AIs, with their ever-increasing sophistication, will penetrate these overly porous boundaries, and your organisation will begin to dissipate.
The good news is that we all possess the means to reinforce our data connectivity. Shared ontologies and URLs as universal identifiers are ready to be deployed, and AI itself can be leveraged to manage the formidable task of achieving comprehensive organisational data integration.
The progress in AI is not just hype; it is very real, but amidst the dazzle of new AI algorithms, I propose a strategic pivot—focus on the unglamorous yet critical task of data integration. Use shared semantics and URLs to make your data cohesive and thus protect your independent identity. The best time to begin integrating your data this way was two decades ago—but the next best time is now!
⭕ Embrace Complexity: https://lnkd.in/eE7fvv3S
⭕ FEP: https://lnkd.in/eSJ3bSFe
⭕ Chris Fields: https://lnkd.in/eG3Kzun2

The idea that AI is cool but that data assets and strategies are king is an emerging race for AI supremacy, as was nicely articulated by Dr. Gevorg Grigoryan, co-founder and CTO of Generate Biomedicine, in this GenengNews article:
“…we are aiming to rapidly build a data-generation capability that will ensure continued differentiation of our ML platform, as we believe algorithms will ultimately be democratized in this space. Furthermore, structural data available publicly, while a great start, do lack in certain important areas, such protein-protein interactions and especially therapeutically relevant antibody-antigen complexes. Importantly, we see cryo-EM as a method that will continue to provide us with the currently missing biological context.”
Sooner or later, AI will inevitably get democratized to the point where pretty much every organization with some meaningful budget will have access to some of the best capabilities on the market (frameworks, out-of-the-box solutions, etc.). Let’s stop fetishizing algorithms and focus on data production, maintenance, and connectivity.
Those who possess the best data assets structured in the right way will stand out and maintain a long-term competitive edge—in the life sciences and in any other industry.
Weekly Highlights
This decision by the Medicines and Healthcare Products Regulatory Agency (MHRA) follows successful clinical trials showing the therapy's effectiveness in significantly reducing symptoms in patients.
Casgevy works by editing the genes encoding hemoglobin in blood-producing stem cells, which are then reinfused into the patient, leading to the production of fetal hemoglobin that alleviates the disease's symptoms.
Expectedly, there are concerns about the high cost and potential unintended genetic modifications by CRISPR-Cas9, even as other countries like the US and those in the EU consider approving the treatment.
From Silicon Valley to the lab of tomorrow: Synfini’s leap to large chemistry models
Synfini, spun out from SRI International, is revolutionizing drug discovery with a multimodal chemistry AI platform, initially developed under DARPA's Make-It program.
Its suite includes SynRoute for synthetic planning, AutoSyn for flow chemistry, SynBuild and SynPlan for molecular design, and SynDB for data.
Synfini's AI models, leveraging pre-trained transformers and neuro-symbolic AI, focus on enhancing medicinal chemists' work by rapidly filtering and prioritizing viable drug candidates.
The platform uses embeddings and machine learning for drug synthesis, emphasizing synthesizability and practical application in automated platforms.
The Singapore-MIT Alliance for Research and Technology has developed a groundbreaking microfluidic method for efficient mesenchymal stem cell (MSC) extraction from bone marrow aspirate, leveraging Deterministic Lateral Displacement (DLD) technology.
This method, detailed in "Lab on a Chip," dramatically enhances MSC yield—doubling it compared to traditional methods—and reduces processing time to 20 minutes.
The DLD-based sorting is label-free, circumventing the complexities and costs associated with conventional fluorescence-activated cell sorting. This innovation streamlines cell therapy manufacturing, improves yield and processing efficiency, and is set to significantly advance stem cell research and therapeutic applications.
DNA writing technologies moving toward synthetic genomes
The field of synthetic biology is advancing rapidly with new DNA writing technologies that are making synthetic DNA more accessible and affordable. Traditional chemical synthesis methods for DNA are being replaced by enzymatic synthesis, which is more efficient and avoids toxic chemicals. This advancement is enabling the synthesis of complex DNA sequences and improving accuracy through optimized assembly and purification processes. Additionally, companies are developing benchtop DNA printers for easier and faster gene synthesis. However, this progress in synthetic genome creation also raises concerns about biosafety, prompting calls for regulatory measures to prevent potential misuse of these powerful technologies.
Race for First Drug Discovered by AI Nears Key Milestone
Insilico Medicine, a biotech firm with bases in Hong Kong and New York, is nearing a significant milestone in AI-driven drug discovery with its experimental lung disease treatment, currently in mid-stage trials in the US and China. This treatment, developed using AI for idiopathic pulmonary fibrosis, represents the first fully AI-based preclinical candidate, marking a potential paradigm shift in drug discovery.
Tune into our Pharma.ai Day on November 24th with follow-up webinars covering feature expansions including the physics and AI-based Alchemistry platform.
ResponderID: the future of RNA biomarkers is now (Genialis)
Integrating spatial transcriptomics data across different conditions, technologies and developmental stages (Nature article)
A landscape of adeno-associated viruses (AAV) production
CLOOME: contrastive learning unlocks bioimaging databases for queries with chemical structures
To Transformers and Beyond: Large Language Models for the Genome
IridescentBio bringing virtual high-throughput assays to antibody R&D
PostEra joins forces with Amgen for AI drug discovery
Label-Free Tracking of Proteins through Plasmon-Enhanced Interference
Generate:Biomedicines opens the hood on its AI software for making proteins
Mark your calendars for the upcoming RNA Leaders Europe Congress, March 13–14, 2024, Basel, Switzerland. The impressive list of speakers includes leaders from Moderna, Alnylam Pharmaceuticals, Johnson & Johnson, Roche, Novo Nordisk, Arrakis Therapeutics, and many others.
Apply the exclusive promo code for our readers, MMBT15, to get 15% off.
Pick of the Week

Noetik Integrates Multimodal Human Data Atlas with In Vivo Functional Genomics Platform for Cancer Immunotherapy Research
San Francisco's biotech scene sees a notable development as Noetik, an AI-native biotechnology firm, integrates its human data atlas with a novel genomics platform, paving a new path in precision cancer immunotherapy.
Central to this advancement is Perturb-Map, a spatial functional genomics technology. Developed by Dr. Maxime Dhainaut, in collaboration with the Icahn School of Medicine at Mount Sinai, Perturb-map facilitates the analysis of hundreds of genetically modified tumor clones in a single experiment. This technology is not just a scientific achievement but a tool that could reshape our understanding of cancer biology.
Noetik’s endeavor focuses on lung cancer, with an initial dataset of over 650 mutations. The application of Perturb-map in this context represents a significant leap in evaluating genetic variants' impacts in vivo, a critical step in developing targeted cancer therapies.
Dr. Dhainaut sheds light on the challenges in functional genomics—the difficulty of conducting gene interrogation at scale within a relevant biological context. Noetik’s technology promises to bridge this gap, allowing in-depth analysis of gene function and interaction within the complex environment of human tissue.
According to Dr. Jacob Rinaldi, CSO & Co-Founder of Noetik, there are limitations of current preclinical models in replicating human cancer biology. However, things are improving with a new paradigm of drug discovery where complex functional genetics and pharmacological hypotheses are tested in vivo, intertwined with advanced machine learning models.
Digital dreams and realities clash pharma and biotech in 2023
This article from Drug Discovery & Development explores the current gap between the AI and digitalization hype and the practical reality of implementing cutting-edge digital technologies in pharma and biotech.
Key takeaways:
The pharmaceutical and biotech industries in 2023 are experiencing a dichotomy between the enthusiasm for AI and digital technologies and the practical challenges of integrating these advancements into existing, often outdated systems. While most companies are devising strategies for AI implementation, with three-quarters planning to do so within two years, the process is complex, involving significant investment, staff training, and workflow modifications.
Despite widespread adoption of digital strategies, many companies struggle with the execution, often delaying major steps due to various challenges like cybersecurity, regulatory concerns, and data connectivity issues. Skepticism about the return on investment for Industry 4.0 initiatives is also notable, particularly in North America.
The shift towards digital data strategies offers significant potential for pharmaceutical and biotech companies, enabling more sophisticated internal use of data for process optimization and decision-making. This transition from traditional, paper-based methods to digital ones promises more comprehensive and effective use of data, leading to improved solutions in drug development.
How AI can help us understand how cells work—and help cure diseases
The Chan Zuckerberg Initiative (CZI) is advancing the concept of a "virtual cell" using AI to model and simulate every cell type in the human body, aiding in understanding diseases and treatment responses. This effort integrates massive datasets, including the developing Human Cell Atlas and OpenCell, to create detailed cell maps, and employs AI models like AlphaFold, ESM, Geneformer, and scGPT for protein structure prediction and gene-cell analysis. CZI is building a high-performance computing cluster with over 1000 H100 GPUs to develop AI models for simulating cellular behavior in health and disease, aiming to democratize access to these resources for global scientific research.
Digging Gems of Biopharma
This week, a visually and factually nice article came from the folks at a16z: Outclassed: The Battle for Therapeutic Market Share.
It explains how, in the past two decades, we've witnessed a remarkable surge in biomedical and therapeutic innovations, highlighted by milestones like the Human Genome Project, CRISPR discovery, and the swift development of COVID-19 mRNA vaccines.
This era, defined by rapid progress in biology, has been driven by substantial increases in investment. Venture capital in the U.S. biotech sector jumped from just under $4 billion in 2000 to nearly $25 billion in 2022, while Pharma R&D investments rose from $38 billion to $83 billion over the same period.
These financial infusions have fostered the creation of diverse new medicinal modalities, such as gene therapy and RNA-based medicine, broadening our capacity to address various diseases, including those previously thought untreatable.
However, as these new modalities emerge, finding their specific applications in treating diseases has become increasingly competitive. The healthcare industry now faces the challenge of integrating these innovations into clinical practice, determining the most effective treatments for specific conditions amid a growing array of options.
Read the article here; it is a must!
Base editing, an innovative gene therapy method, was found to significantly reduce "bad" cholesterol in clinical trials
31% of all global deaths are due to cardiovascular diseases, and bad cholesterol is among the key risk factors here.
Bad cholesterol, known as LDL (low-density lipoprotein), contributes to plaque buildup in the arteries, increasing the risk of heart attacks and strokes. Effectively managing LDL levels is crucial to combating this prevalent health threat.
While in many cases, you can reduce LDL simply by eating healthy and doing regular exercises, many people have genetic causes of high LDL, and they must take statins every day for their entire lives. Which is obviously hard to maintain. Also, statins are not helpful enough in some cases.
Imagine an alternative way: you just do one genetic manipulation and forget about statins for decades? That's what Boston-based Verve Therapeutics tries to achieve.
Early clinical trial results are promising!
In a pivotal trial, Verve Therapeutics used base editing, a CRISPR-derived gene therapy, to effectively reduce bad cholesterol by targeting the liver's PCSK9 protein. This more precise method alters a single DNA strand, differing from traditional CRISPR. It was tested on individuals with heterozygous familial hypercholesterolemia (FH), aiming to enhance LDL receptor activity for cholesterol clearance. High-dose patients showed notable LDL and PCSK9 reduction.
To be fair, safety concerns still remain to be addressed.
Following its early-stage trial data release, Verve Therapeutics' stock fell quite significantly, against the fears that the new therapy VERVE-101 didn't notably outperform Novartis AG's Leqvio.
Despite concerns over transient liver enzyme elevations in some participants, experts remain hopeful about VERVE-101's potential in reducing LDL cholesterol in patients with familial hypercholesterolemia. Now stocks are going up again.
I am very optimistic about this trial and the advent of in vivo base editing, too (I do not hold their stocks, but I do have issues with managing LDL using just diet and regular sport...).
This news comes after Eli Lilly & Co. acquired rights to some of Verve's gene-editing treatments.