BioByte 073: citizen science for DNA sequence alignment, cellular niches, updates in AI antibody design, polarizing biotech investment trends
Welcome to Decoding Bio, a writing collective focused on the latest scientific advancements, news, and people building at the intersection of tech x bio. Happy decoding!
What we read
Blogs
We Need to be Ready for Biotech’s ChatGPT Moment [Eric Schmidt, Time, April 2024]
In a recent article by Eric Schmidt, the biotech revolution is portrayed as an imminent transformation poised to redefine industries much like the digital revolution did. Schmidt, a key figure in emerging biotechnologies, describes an era where biomass-derived materials, personalized therapies, and lab-grown food could vastly improve how we manage health, environment, and food security.
2023 marked notable advancements: the U.S. green-lit lab-grown meat sales, and AI made strides in predicting new materials' structures potentially useful in electronics. The CRISPR gene-editing tool Casgevy also received approval, showcasing the speed of biotechnological advancements.
Schmidt argues that integrating AI with biotech will accelerate our capability to innovate, akin to how ChatGPT revolutionized data handling. This synergy could lead to efficient disease combat, enhanced energy production, and precise carbon capture.
However, maintaining leadership in biotech necessitates a robust domestic supply chain and substantial investment in human and physical capital. The current U.S. administration has initiated measures to protect sensitive genetic data from foreign adversaries, highlighting the security implications tied to biotech.
Guardrails are essential to prevent misuse of biotechnological advancements, with structured red-teaming recommended to identify vulnerabilities preemptively. Public-private collaborations are deemed crucial to foster a secure and innovative biotech ecosystem.
Chemists invent drugs and save lives [Donald Weaver, ChemMedChem, April 2024]
Dr. Donald Weaver highlights the underappreciated role that chemists play in inventing drugs in a beautiful editorial that starts his own experiences as a physician treating dementia and Alzheimer’s Disease. He describes the long history of chemists creating transformative drugs like aspirin, barbiturates, and other medicines that have alleviated suffering across diseases and notes that progress has been significantly slower for neurological disorders like Alzheimer's. Despite many failed attempts so far, Weaver remains optimistic that emerging techniques in medicinal chemistry, like computer-aided drug design, high-throughput screening, and a deeper understanding of disease mechanisms, will eventually lead to effective Alzheimer's treatments and beyond. He underscores the underappreciated role of chemists, not physicians, in discovering and synthesizing new drugs and how there’s a need for better communication highlighting chemists' stories and role in drug development, which could ultimately lead to greater support and resources. While the message is simple, his inspiring storytelling is worth a read.
Biotech fundraising in 2024: a story of haves and have-nots [Rowan Walrath, C&EN, April 2024]
In 2024, biotech and pharmaceutical companies are experiencing a significant shift in fundraising dynamics, as outlined by Rowan Walrath in Chemical & Engineering News. Despite an overall increase in the capital raised, totaling $5.9 billion in the first quarter, the number of deals has dropped to the lowest since the third quarter of 2018, with only 209 recorded transactions.
This year, investment trends are polarized; fewer companies are managing to secure funding, but those that do are attracting substantial investments, often surpassing $100 million per funding round. This trend underscores a growing preference among investors for backing ventures with more mature drug candidates and proven management teams, minimizing perceived risks.
Companies like Mirador Therapeutics, which launched with a robust $400 million thanks to its experienced founders, and Synnovation Therapeutics, with a significant $102 million initial funding, exemplify this trend. Investors are favoring these "mega-rounds" to concentrate resources on fewer, but potentially more promising, bets.
Moreover, there is a shift towards funding scientifically validated approaches, with companies like Cour Pharmaceuticals and Latigo Biotherapeutics successfully raising funds based on the clinical progress of their drug candidates.
However, this investment strategy may sideline early-stage innovation, potentially stifling groundbreaking new technologies and treatments. The concentration of capital into established entities with near-term potential for success poses challenges for newer, unproven biotech ventures that might hold transformative scientific proposals.
Academic papers
Improving microbial phylogeny with citizen science within a mass-market video game [Sarrazin-Gendron et al., Nature Biotech, April 2024]
Why it matters: A clever use of citizen-science in a mass market video game by mobilizing over 4 million gamers to participate in DNA sequence alignment, resulting in superior accuracy and performance as compared to state-of-the-art computational methods.
This paper underscores the significant potential of integrating scientific research with popular gaming platforms, demonstrating how gamification can transform traditional data processing tasks into engaging public activities. By mobilizing over 4 million gamers to participate in DNA sequence alignment through a video game, the study highlights an innovative approach to crowd-sourcing complex scientific data analysis. This method not only accelerates scientific discoveries by leveraging a vast pool of untapped cognitive resources but also enhances public understanding and engagement with science. Consequently, this approach could serve as a model for future scientific endeavors, promoting wider public participation in research and potentially leading to breakthroughs in various fields by harnessing the collective power of global gaming communities.
In one of the more impressive applications of citizen science we’ve seen, a paper out in Nature Biotech this week describes the design of a mini-game called Borderlands Science within Borderlands 3 (a massively successful video game with millions of players), which tasks players with aligning sequences of microbial DNA drawn from the human microbiome (DNA sequence alignment is the process of arranging DNA sequences to identify regions of similarity and differences, which helps in understanding evolutionary relationships, gene functions, and genetic variations). Players align these sequences by solving tile-matching puzzles, which represent the DNA sequences as colored bricks that need to be arranged in a specific order. Over 4 million players have participated, collectively solving more than 135 million puzzles and significantly contributing to the alignment of 1 million 16S ribosomal RNA sequences.
The results from these crowd-sourced alignments showed improvements in microbial phylogeny estimations over traditional computational alignment methods. Notably, the alignments produced through Borderlands Science enhanced the UniFrac effect sizes for various microbiome-related variables, indicating better resolution in distinguishing microbial communities associated with different human health conditions.
Capturing and modeling cellular niches from dissociated single-cell and spatial data [Nature Biotechnology, 2024]
Why it matters: High resolution spatial tools are enabling scientists to increase their understanding of how cells interact with their neighbors and the surrounding tissue. Existing methods typically simplify the complex interplay within cellular environments by reducing data to mere cell-type fractions.
To address this, the researchers developed COVET, which models the microenvironment by capturing gene-gene expression variations among neighboring cells. This approach not only reflects biological relationships but also minimizes the impact of measurement artifacts.
Further enhancing this field, the paper introduces ENVI, a machine learning framework that integrates the niche information derived from COVET with both spatial and single-cell data. This integration is adept at imputing missing genes and mapping spatial data directly onto dissociated single cells. ENVI has proven effective across various platforms, revealing spatial expression trends and localizing distinct cell types in complex tissues, thereby surpassing existing methods.
Fig. 1: COVET represents the cellular niche for spatial analysis and for spatial inference of scRNA-seq data.
Atomically accurate de novo design of single-domain antibodies [Bennett et al., bioRxiv, 2024]
Why it matters: de novo computational design of structurally accurate antibodies has remained elusive. Given the diversity and quality of interfaces that the diffusion model RFdiffusion can design, it should be possible to develop fine tuned versions to design de novo antibodies. The authors did just that to design VHHs against disease-relevant targets and confirmed atomically accurate structures through cryoEM. There is, of course, room for improvements as the affinities and success rates were low.
In this preprint published last month by the Baker lab, the authors finetune RFdiffusion and RoseTTAFold2 for antibody design and design validation respectively.
They initially focused on designing single-domain antibodies (VHHs) based on the variable domain from heavy-chain antibodies. So far two VHH-based therapies are approved by the FDA, with many in clinical trials. The average interaction surface area of a VHH is similar to that of a variable fragment (Fv), suggesting a method that could be suitable for Fv design as well. Designs against the targets were screened either at high-throughput by yeast surface display or low-throughout with E. coli expression and single-concentration surface plasmon resonance.
RFdiffusion was successful at generating moderate affinity VHHs against diverse epitopes of disease-relevant targets. To evaluate the design accuracy, the anti-HA VHHs in complex were subjected to cryo-EM structure determination, which showed that the de novo designed VHH accurately bound a natively glycosylated epitope with atomic-level precision.
Notable Deals
Vertex Enters Into Agreement to Acquire Alpine Immune Sciences for $4.9B in cash
Nurix Therapeutics Announces Pricing of Upsized $175.0 Million Public Offering
Regeneron enters venture investing with $500M, poaches senior partner from ARCH
Events
Mila is organizing a Machine Learning for Drug Discovery Summer School from June 12th - 21st. The program includes 5 days of lectures from professors like Yoshua Bengio, Michael Bronstein, Anne Carpenter, and more, followed by the Molecular Machine Learning Conference and a 2-day hackathon. If you're a data scientist/ML engineer or a scientist with a programming background, this is a great opportunity to get hands-on training and learn how to apply ML to drug discovery.
Early bird tickets are available now! Register here.
Field Trip
Did we miss anything? Would you like to contribute to Decoding Bio by writing a guest post? Drop us a note here or chat with us on Twitter: @ameekapadia @ketanyerneni @morgancheatham @pablolubroth @patricksmalone