Weekly Tech+Bio Highlights #1
Building Foundation Models for Cancer Biology; New General Purpose LLM for Chemical and Biomedical Applications; Owkin Unveils AI-Driven Pipeline; Quantum Computing in Biopharma, & more.
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!
Weekly tech+bio highlights
🚀 AstraZeneca plans to build a $1.5 billion manufacturing facility in Singapore for antibody-drug conjugates (ADCs), marking its first end-to-end ADC production site and supported by the Singapore Economic Development Board. The facility, set to be operational by 2029, aims for zero carbon emissions from day one.
🔬 Insilico Medicine, in collaboration with NVIDIA, introduces nach0, a large language model (LLM) for solving biological and chemical tasks, integrating diverse datasets to perform natural language processing and molecular generation, as detailed in Chemical Science Journal.
💰 LabGenius raises £35M in Series B led by M Ventures to advance its ML-driven antibody discovery platform and progress a pipeline of assets towards clinical development, bringing total funding to £58M.
🔬 Sanofi, Formation Bio, and OpenAI announce a first-in-class AI collaboration to accelerate drug development, leveraging proprietary data, AI model customizations, and extensive engineering resources to bring new medicines to market more efficiently.
📈 Tempus AI has filed its S-1 to go public, highlighting its extensive multimodal, de-identified records database, which is over 50 times the size of The Cancer Genome Atlas.
✨ Pharma giants like Merck, J&J, Roche, and Amgen filed quantum computing patents, while Pfizer and IBM collaborate to integrate generative AI and quantum computing to enhance clinical trials.
🔬 Owkin expands its oncology and immunology drug pipeline with the global licensing of OKN4395, a dual EP2/EP4 inhibitor from Idorsia, leveraging advanced AI and multimodal patient data for precision medicine.
💰 Orna Therapeutics acquires ReNAgade Therapeutics, appointing Amit Munshi as CEO and retaining Tom Barnes on the board. The acquisition combines Orna's circular RNA technology with ReNAgade's RNA delivery systems to advance in vivo CAR RNA medicines for oncology and autoimmune diseases.
🚀 Radar Therapeutics launches with $13M to develop RNA sensor programmable technologies, aiming to deliver genetic payloads that activate only in the presence of specific cellular expression signatures.
Building Foundation Models for Cancer Biology
NOETIK team are training self supervised world models that can learn cancer biology from primary data. These models are queryable, enabling users to ask questions to discover new therapeutic insights. In the example below, the team recapitulates some fundamental biology around PD-L1 and T cells.
NOETIK is an AI-native biotechnology company founded in 2022 in San Francisco, California.
The company focuses on leveraging advanced machine learning and self-supervised learning to discover and develop precision cancer immunotherapies by analyzing multimodal human data, including genomics, transcriptomics, and proteomics. Recently, Noetik raised $14 million in seed financing to support its data generation efforts and announced the launch of its "Perturb-map" platform for in vivo functional genomics, aimed at accelerating the discovery of new cancer drug targets
These are driven by massive, proprietary, multimodal human patient data generated in their own labs. They are profiling tumors at a scale and depth never before achievable: protein, RNA, DNA, H&E. All of these data feed 500 M parameter multimodal models trained on 100s of NVIDIA H100 GPUs over the course of weeks.
As Dr. Ron Alfa, Co-founder and CEO at NOETIK.ai explains in his LinkedIn post:
“ …these models are scalable — this is just the beginning. We're pointing this engine at one of the most challenging problems in drug discovery, finding the right drugs for the right patients.”
Meet nach0: a New General Purpose LLM for Chemical and Biomedical Applications
Insilico Medicine, a clinical-stage AI-driven drug discovery company, in collaboration with NVIDIA, has introduced nach0, a novel large language model (LLM) transformer designed for biological and chemical tasks.
As detailed in a recent publication in the Chemical Science Journal, nach0 is distinguished by its ability to handle multi-domain and multi-task applications, such as natural language understanding, synthetic route prediction, and molecular generation.
Existing biomedical LLMs like BioBERT and SciFive primarily focus on biomedical text mining without incorporating chemical structure descriptions.
Although models such as Galactica integrate both text and chemical structures, they lack training for diverse chemical tasks.
Nach0 addresses this gap by utilizing a comprehensive dataset that includes abstract texts from PubMed and patent descriptions from the U.S. Patent and Trademark Office, totaling 100 million documents.
These were converted into 355 million tokens from abstracts, 2.9 billion tokens from patents, and 4.7 billion tokens representing molecular structures in the simplified molecular-input line-entry system (SMILES).
Owkin Unveils AI-Driven Oncology and Immunology Pipeline
Owkin has unveiled its AI-driven precision drug pipeline in oncology and immunology, highlighting the in-licensing of OKN4395 from Idorsia.
OKN4395, a potent EP2/4 dual antagonist, is set to enter Phase 1 trials in early 2025, and is anticipated to be best-in-class due to extensive medicinal chemistry efforts. The preclinical data for OKN4395 shows promising restoration of T cell functionality.
Owkin's pipeline features four early-stage de novo drug discovery programs alongside this patent-protected asset. The company leverages its proprietary AI engines for biomarker discovery and drug positioning, optimizing clinical trials and patient selection.
These AI tools utilize multimodal patient data, including the world's largest spatial multiomics dataset, MOSAIC, to capture the tumor microenvironment.
One of the central features of OWKIN’s platform is the idea of federated learning, which is realized via OWKIN Connect, which links multiple partners on a single interface.
Federated learning enables decentralized training of machine learning models by processing data locally and aggregating model updates centrally, ensuring data privacy.
Data scientists train AI models locally on their data without transferring it, then share the trained model parameters with a central aggregator. This central aggregator, either a partner center or Owkin, averages these parameters to create a more robust global model.
This iterative process, repeated multiple times, refines the model, making it more predictive and generalizable.
Federated learning is particularly advantageous for institutions like hospitals that handle sensitive data, allowing them to collaborate without compromising patient confidentiality.
According to press release, Owkin plans to further expand its pipeline by combining in-house discovery with additional in-licensed assets in oncology, immunology, and inflammation.
This TechBio Company To Unveil Novel Cell Analytics Technology
Cytomos, a Scotland-based TechBio company, is set to unveil its advanced cell analytics technology, Celledonia™, at the ISCT 2024 annual meeting in Vancouver.
This new platform, built on the patented AuraCyt™ technology, promises to enhance cell and gene therapy (CGT) processes by offering label-free, real-time cell monitoring and analytics. The technology aims to improve decision-making and process control, accelerating the development and manufacturing of complex biologics.
Pioneering: AuraCyt™ accelerates biological drug discovery and development, and streamlines manufacturing, with near-real-time monitoring potential and better prospects for automation
Label-free: The technology harnesses the power of dielectric spectroscopy and microelectronics and characterises cells without the need for labels and directly in media
Unbiased: The technology addresses the analytical challenges of cell and gene therapy developers and monoclonal antibodies (mAbs), with fast and sensitive detection of single-cell attributes. The platform generates predictive analytics for consistent decision-making without operator bias
Uniquely scalable: AuraCyt™ is the only scalable technology that can measure cellular physiology, based on near-real-time intrinsic single-cell properties
Quantum Computing in Biopharma: A Quantum Leap Forward
Pharmaceutical and healthcare companies are increasingly investing in quantum computing, anticipating its potential to revolutionize drug design, clinical trials, and personalized medicine. The promise of quantum computing lies in its unparalleled speed and capability to process complex datasets, making it a game-changer for the biopharma industry.
Key Companies and Initiatives
Merck, Johnson & Johnson, Roche, and Amgen: In 2023, these pharma giants filed patents related to quantum computing, signaling their commitment to leveraging this technology.
Moderna and IBM: Partnered to explore the integration of quantum computing and generative AI for drug development.
AstraZeneca and Sanofi with SandboxAQ: Collaborated with Alphabet's spinoff to advance quantum computing applications in healthcare.
Pfizer and IBM: Formed a collaboration to use quantum computing for improving clinical trial performance and predicting trial site failures.
Boehringer Ingelheim: Employing quantum algorithms to study small molecule behavior and predict protein-ligand binding energies.
Novo Nordisk Foundation (NNF): Investing $200 million in building a quantum computer at the University of Copenhagen, along with a $212 million commitment to develop a generally applicable quantum computer through a 12-year program.
Quantinuum (formed by the merger of Cambridge Quantum and Honeywell Quantum Solutions): Developing quantum tools for drug discovery, with industry partners including Roche, Amgen, and AbbVie.
QC Ware and IBM: Collaborated with Roche to demonstrate that quantum simulations can outperform classical computers in detecting diabetic retinopathy.
QuantumBasel: Hosting quantum computers from IBM, D-Wave Systems, and IonQ, with a focus on democratizing access to quantum computing and fostering innovation in health applications.
Algorithmiq: Utilizing the IBM Quantum System One at Cleveland Clinic, combined with its quantum AI software Aurora, to develop light-activated cancer therapies.
Quantum Advantage in Biopharma
Quantum computers use qubits, which can exist in multiple states simultaneously, unlike classical computers that use binary bits. This property allows quantum computers to process vast amounts of data and provide multiple potential outcomes, enabling:
Molecular Dynamics Simulation: Quantum computing can model complex molecular interactions, such as protein folding and ligand-protein binding, more efficiently than classical computers.
Clinical Trial Optimization: Quantum algorithms can analyze patient data to optimize clinical trial design and predict recruitment outcomes, reducing trial durations and improving precision.
Drug Discovery Acceleration: By integrating genomics, transcriptomics, and other 'omics' data, quantum computing can enhance predictive modeling for drug targets and accelerate the discovery process.
Challenges and Future Outlook
While the potential of quantum computing in biopharma is immense, several challenges remain:
Hardware Limitations: Current quantum computers are still in their infancy and require advancements to become fault-tolerant and scalable.
Investment Uncertainty: Despite significant interest, venture capital funding for later-stage quantum computing applications has been challenging to secure.
Talent Shortage: There is a paucity of scientists trained in quantum technologies, which could impede the field's growth.
Regulatory and Data Privacy Concerns: Ensuring the confidentiality and security of patient data processed by quantum computers is critical.
Despite these hurdles, the industry is optimistic about the future, with expectations that quantum computing will begin delivering tangible benefits by 2030. As hardware and algorithms advance, the biopharma sector is poised to harness the full potential of quantum computing, marking a significant leap forward in healthcare innovation.
New Discovery Potentially to Re-write Biology Textbooks
The new article discusses a groundbreaking discovery in molecular biology, where scientists have found a unique bacterial enzyme that creates new genes by reading RNA templates and converting them into DNA. This defies the traditional understanding of genetic information flow, which typically moves from DNA to RNA to proteins. The key points of the article are:
Reverse Transcriptases and Their Role: Traditionally, reverse transcriptases are enzymes that viruses use to transcribe RNA into DNA, contrary to the usual DNA to RNA transcription process.
New Bacterial Mechanism: Researchers discovered a bacterial version of reverse transcriptase that reads RNA to create new DNA sequences, which then encode for new genes. This newly discovered process was found in the bacteria Klebsiella pneumoniae.
Mystery RNA and the Neo Gene: The bacterial defense system includes DNA for a reverse transcriptase and a mysterious RNA sequence. This system generates long DNA sequences with repeated segments, matching the mysterious RNA and forming what the researchers called the "never-ending open reading frame" (neo). This sequence lacks a typical stop signal, making it theoretically endless.
Protein Production and Function: Upon viral infection, the neo sequence is transcribed into RNA and translated into the Neo protein, which inhibits cell division. The exact mechanism by which Neo halts cell growth remains unclear, but the protein appears to form helical structures.
Scientific Implications: This discovery challenges existing paradigms of genetic information flow and opens new avenues for biotechnological applications. The findings, though not yet peer-reviewed, suggest that reverse transcriptase in bacteria can generate entirely new genes, which could have significant implications for genetic research and biotechnology.
Expert Reactions: The scientific community has reacted with astonishment, likening the discovery to something out of alien biology. Researchers see potential for future applications and believe this could fundamentally alter the understanding of genomes.
The discovery was co-led by molecular biologist Stephen Tang and biochemist Samuel Sternberg, and the findings were posted on the bioRxiv preprint server.
Some Topical Newsletters You May Like:
Using Quantum-Enhanced AI to Design Cancer Drugs For the First Time
AI Foundation Models in Biotech: New Paradigm
How Industry Embraces Organ-on-Chips: A 2024 Status Report
A Landscape of Novel Antibody-Drug Conjugates
11 Biopharma Trends to Watch in 2024