Friends and colleagues often ask, “What are the top 100 important "AI in Biology" papers that provide broad insights into the field?” 📚🔬 While narrowing it down to exactly 100 is no small feat, I’ve curated a list of foundational and impactful #BioAI papers. I'm sure the list would exceed more than 100 given the relentless expansion of this thrilling field! Please follow this thread for key insights, and please feel free to suggest any papers I may have missed! With my background and interests, I prioritize papers in #AgingBiology, #CellBiology, and #Neuroscience. I’ll keep this thread pinned to my profile and update it regularly. I encourage everyone to add more papers from their own expertise—let’s make this interactive and foster engaging discussions! Here is a collection of essential #AIbio papers. #100_AIBio_Papers #AIBio_papers #AIBio_Chat #AIneuro_papers #AI_AgingBio_papers

Transferable generative models bridge femtosecond to nanosecond time-step molecular dynamics science.org/doi/10.1126/scia… AI Speeds Up Molecular Simulations by 10,000× A new AI framework called TITO represents a significant advance in computational chemistry and biophysics. The system was trained on more than 12,500 organic molecules and over 1,000 peptides, learning the fundamental rules that govern how molecular systems evolve over time. Traditionally, molecular dynamics (MD) simulations track atomic movements using femtosecond-scale time steps (10⁻¹⁵ seconds). While highly accurate, this approach is computationally expensive and often prevents researchers from observing the slower molecular motions—such as protein folding, conformational changes, and relaxation processes—that determine biological and chemical function. TITO addresses this challenge by using a deep generative modeling approach that can accelerate molecular simulations by up to four orders of magnitude (10,000×) while preserving the underlying physics of the system. Rather than calculating every individual atomic step, the AI learns how molecular structures evolve across time and can effectively bridge molecular behavior from femtoseconds to nanoseconds. Remarkably, the model not only reproduces equilibrium molecular ensembles and dynamical relaxation processes but also generalizes across different chemical compositions and molecular sizes. It can even extrapolate to peptides larger than those used during training, capturing chemically meaningful transitions on timescales previously accessible only through costly brute-force simulations. Why This Matters This work could dramatically accelerate research in: Drug discovery — exploring protein-ligand interactions and conformational states more efficiently. Materials science — designing novel materials with desired properties. Catalysis — understanding reaction pathways and optimizing catalysts. Biophysics — studying protein dynamics, folding, and molecular mechanisms at unprecedented speed. More broadly, TITO may represent one of the clearest examples of AI accelerating scientific discovery itself, not merely assisting with data analysis. By expanding the accessible timescales of atomistic simulations while maintaining physical realism, AI-driven approaches like TITO could transform how scientists investigate molecular systems, enabling discoveries that were previously impractical due to computational limitations. For researchers working at the intersection of AI, chemistry, and biology, this development highlights a growing trend: AI is increasingly becoming a tool for exploring the laws of nature, not just interpreting experimental data. The ability to rapidly map molecular conformational landscapes, thermodynamics, and kinetics may ultimately shorten the path from hypothesis to discovery across multiple scientific disciplines. #AIforScience #MolecularDynamics #DrugDiscovery #GenerativeAI #Biotech

Navigating uncertainty: It notes that recent advances make human-level AGI seem more concretely imaginable than before, and continued trends such as improved algorithmic efficiency, large-scale investment in compute, and AI infrastructure expansion could plausibly sustain rapid growth through the end of the decade. Even challenges like limited high-quality training data might be mitigated using self-generated data from reinforcement learning, simulations, and agent-based systems. However, it emphasizes that AI progress is inherently unpredictable. Because growth may be exponential or even hyperbolic, small uncertainties in assumptions can lead to large differences in outcomes. As a result, average or “best guess” forecasts may be misleading for decision-making. Progress could either continue rapidly or hit natural limits sooner than expected, potentially delaying AGI. The text also stresses that even if AGI is achieved or surpassed, this does not guarantee “omnipotent” ASI or breakthroughs like curing aging, nanotechnology-based matter control, or climate restoration—such outcomes are not guaranteed. A central tension is between accelerating forces (like AI helping accelerate research) and decelerating forces (like increasing difficulty and resource requirements in science). These opposing dynamics may coexist and compete, making it hard to determine which will dominate long-term. Because of this, reliable forecasting requires using multiple models, tracking quantitative indicators (like algorithmic efficiency), and continuously updating predictions rather than relying on a single estimate. #omnipotent #Cure_Aging

2/2 Is the Singularity near? Evidence from scaling laws and recent benchmarks suggests that, at least so far, capability growth has often been super-linear with compute, though not perfectly consistent. Importantly, the text emphasizes that even if individual model capability plateaus, overall system capability could still grow through scaling deployment: running many more instances of models, running them faster, or letting them “think” longer. This “quantitative scaling” might produce effects that resemble qualitative breakthroughs. The example given is that if AGI were achieved but expensive, continued 10× annual growth in compute could rapidly expand usage from thousands to millions of instances, raising the question of whether such scaling alone could eventually produce ASI. #Singularity #AGI #Scaling #compitational_power

1/2 Is the Singularity near? The passage discusses uncertainty about whether AI progress could lead to a “singularity” driven by continued growth in effective compute. It argues that even if traditional drivers like investment or hardware improvements slow down, improvements in training efficiency and system optimization could still sustain exponential growth in usable compute (e.g., 10× per year). This would enable larger training runs, cheaper model deployment, and more intensive “test-time compute” (like reasoning, planning, or multi-agent systems). A key uncertainty is how this compute growth translates into new AI capabilities. Two possibilities are raised: 1. Diminishing returns could mean ever-increasing compute is needed just for small capability gains → slow progress. 2. Or capabilities could scale proportionally with compute → continued exponential improvement.

Single-cell proteome atlas of aging mouse microglia reveals subpopulation-specific phagoproteome cell.com/neuron/abstract/S08… @NeuroCellPress

A strong sustainability approach to AI development | Nature Machine Intelligence nature.com/articles/s42256-0…

scientist feel pressured to use AI tool? 😀

Many people are currently unable to submit their abstracts due to website errors this afternoon. Is @SfNtweets considering an extension of the deadline this year as well, given the technical issues?

This is definitely the beginning of a STEM renaissance. It makes me wonder: how many modern-day Da Vincis will emerge from this new era of scientific and technological discovery?

What books are AI researchers, founders, and scientists reading these days? I'm interested in books that go beyond models and benchmarks and tackle bigger questions: • Intelligence and AGI • Neuroscience • Longevity • The future of science Recommendations are appreciated ? Here are the books I’m currently studying and will continue with #AIBook #AIreading_list #AGI #Neurosciences #Longevity

Indeed, next decade will be a scientific renaissance for people who know how to use existing tools and methods . After reading these kind of posts, I always wish I have even more time to upgrade myself and read more AI stuff! Time to start a reading club, I guess . Feel share your interest to participate in!!

What would a Figma for science look like? Drug discovery reasoning is inherently spatial and non-linear. A canvas lets the work breathe. Humans will remain in the AI x bio loop for the foreseeable future, and some of the most exciting work is happening at the UX layer.

The problem ASA solves is not one bad answer. x.com/Symbioza2025/status/20… It is the slow movement of a system away from its original intent while everything still appears coherent.