Faster Science, Better Drugs

The Core Problem: Why Science is Slow

Scientific progress, particularly in biology, is hindered by a “weird Gordian knot” of factors. Unlike AI research that moves at the speed of GPUs, biological research involves moving atoms and is constrained by the real-time processes of growing cells, tissues, and animals. The core issues are multifactorial, stemming from incentives, funding structures, and the training system.

A significant challenge is the increasing need for interdisciplinary collaboration. It is exceptionally difficult for individual research groups or companies to excel at more than two distinct domains simultaneously (e.g., computational biology and genomics). Modern problems require integrating five or more fields, such as neuroscience, immunology, machine learning, chemical biology, and genomics.

The Arc Institute’s Approach: Fostering Collaboration

The Arc Institute was founded as an “organizational experiment” to address these challenges. By bringing experts from five distinct domains under one physical roof, the goal is to “increase the collision frequency” and unlock a new space of research problems. This model contrasts with a traditional university setting, where physical distance and misaligned incentives often discourage deep collaboration. In academia, researchers are primarily incentivized to publish their own papers and make their own discoveries, not necessarily to work on larger, collective flagship projects. Arc is designed to enable these larger projects, such as finding new Alzheimer’s drug targets and building “virtual cells.”

The Moonshot: Simulating Biology with “Virtual Cells”

The central moonshot at Arc is to create “virtual cells” to simulate human biology using foundation models. The ambition is to make these models the default tool for experimentalists, accelerating discovery to the speed of a neural network’s forward pass.

However, modeling biology with AI is fundamentally harder than modeling language or images.

• Lack of Native Intuition: Humans are native speakers of language and interpreters of images, making it easy to evaluate the output of models like GPT-4 or DALL-E. In contrast, we don’t “speak the language of biology” and can only interpret it with a “thick accent.” Evaluating a DNA foundation model’s output is not intuitive.
• The “Lab in the Loop” Bottleneck: The iteration cycle for biological models is slow because predictions must be validated with physical lab experiments. Increasing the speed and dimensionality of this experimental feedback is a critical challenge.
• Incomplete Data: We are almost certainly not measuring all the critical components of a cell. While we can scale the measurement of transcriptional information (RNA), this is only a “lower resolution mirror” for what is happening at the protein or metabolic level. The strategy is to bet on what can be scaled today (genomics) and layer in other data modalities over time, trusting that scaling laws will help fill in the gaps.

Defining the Virtual Cell: Perturbation Prediction as the Goal

The most famous success of ML in biology is AlphaFold, which accurately predicts a protein’s 3D structure from its amino acid sequence. The goal for virtual cells is to achieve a similar “AlphaFold moment” for cell biology.

At Arc, this is operationalized as perturbation prediction. The model’s core task is to predict the necessary interventions to move a cell from one state to another across a manifold of cell states (e.g., from inflamed to quiescent, or from a fibroblast to a stem cell). This directly mirrors the process of drug discovery, which is fundamentally about finding a molecule (a perturbation) that shifts a cell from a disease state to a healthy one.

The objective is to create a practical “co-pilot for a wet lab biologist” that can suggest combinatorial perturbations and facilitate in-silico target identification, ultimately forming the basis for a vertically integrated, AI-enabled pharma company.

The Path to an “AlphaFold Moment” for Cell Biology

Progress is framed in terms of GPT-like capabilities. The field is currently somewhere between GPT-1 and GPT-2. The “blurry pictures of life” generated by current models are not yet alive, but a clear path for improvement exists through scaling.

A “GPT-3 moment” for virtual cells would be when the model can recapitulate canonical biological discoveries from first principles. For example:

• Can the model predict that the four Yamanaka factors are required to reprogram a fibroblast into a stem cell?
• Can it rediscover differentiation factors that turn a stem cell into a neuron or muscle cell?
• Can it predict the mechanism of action of existing FDA-approved drugs?

Achieving these milestones would provide tangible, understandable benchmarks that move beyond abstract ML metrics like mean absolute error, making the model’s capabilities clear even to biologists who have “never touched a terminal.”

Bridging AI Breakthroughs with Industry Impact

While AI promises to revolutionize discovery, the biotech and pharma industries face persistent challenges. 90% of drugs fail in clinical trials, either because the wrong biological target was chosen or the drug composition was ineffective.

Even with a 90% accurate virtual cell, new challenges will arise. The model might suggest a highly specific intervention, like “target this GPCR only in the heart but not in any other tissue,” for which no corresponding drug technology currently exists. This highlights how progress in one area creates new research problems in another, reinforcing the need for an integrated research institute.

The recent success of GLP-1 drugs has added over a trillion dollars in market cap to companies like Eli Lilly and Novo Nordisk. This demonstrates the immense value created by tackling diseases affecting large patient populations and has, in turn, “net increased the ambition of the industry.”

The Unsolved Bottlenecks: From Lab to Clinic

Despite advances in AI-driven design, the physical world remains the ultimate bottleneck.

“Even if we can solve the designing part… we still need to make these things physically and test them in animals and hopefully predictive models and then actually in people. And I think, you know, that will increasingly be the bottleneck in many ways.”

The major hurdles that slow down the translation of science into medicine are:

1. Capital Intensity: Drug development is incredibly expensive. Better models are needed to increase success rates and thereby lower the effective capital intensity.
2. Clinical Development: The process of running human clinical trials is slow and hard to compress. While technology can help optimize trial design and enrollment, proving a survival benefit for a cancer drug or a longevity benefit for an aging drug inherently takes time.
3. Manufacturing and Testing: Molecules designed in-silico must be synthesized and validated through a long chain of experiments: in mice, mutts, monkeys, and finally, man.

Hype, Hope, and Heft in AI for Drug Discovery

When evaluating AI’s current impact on drug discovery:

• Hype: Toxicity prediction models and vague “multimodal biological models.”
• Heft (Real Impact Today): Anything involving proteins (binding prediction, protein design) and pathology AI for automating the work of pathologists and radiologists. In the near future, AI will cease to be a novelty and will become a native part of the entire R&D stack, integrated into every step of the process.

• Tags:
• Foundation Models
• Drug Discovery
• Computational Biology
• AlphaFold
• Virtual Cells
• Arc Institute