The Infrastructure Layer: 30 Platforms Powering Human-Relevant Drug Development

In parallel, over 100 researchers working under the MICrONS program released a map of one cubic millimeter of mouse brain tissue. The team, led by the Allen Institute, Princeton, and Baylor College of Medicine, reconstructed 200,000 cells and 523 million synapses using 28,000 brain slices. The resulting dataset is 1.6 petabytes of high-resolution imaging paired with in vivo activity recordings.

Built through 20 years of effort, the MICrONS Project reveals the most detailed 3D map of a mouse brain region to date, combining real-time neural activity with synapse-level structural detail. Image: Allen Institute, Tyler Sloan in Nature

It’s less than 1% of a mouse brain. But it already revealed new circuit motifs involving inhibitory neuron types like chandelier and Martinotti cells. The dataset is open access and intended as a scaffold for future precision neurotherapeutics.

Looking back to 1965, Margaret Dayhoff compiled just 65 protein sequences in her Atlas of Protein Sequence and Structure. The entire dataset fits into a few kilobytes—small enough to store on a floppy disk or email as a modern PDF. Today, a single biomedical project can generate petabytes. That’s a billion-fold increase in volume.

This kind of output, unthinkable two decades ago, is now routine in many areas of biomedicine. GenBank stores over 250 billion nucleotide bases. UK Biobank holds more than 30 petabytes across genomic, imaging, and health data. Tempus AI processes 300 petabytes across multimodal clinical and molecular datasets. AlphaFold’s 200 million protein predictions add another 23 terabytes of modeled biology.

What made this scale-up possible? Faster compute, cheaper storage and better compression. And a growing ecosystem of tools to ingest, clean, and analyze multi-source biomedical data in real time.

In a major policy shift, the FDA had just expanded the use of New Approach Methodologies (NAMs) in Investigational New Drug (IND) submissions, starting with monoclonal antibodies and other therapeutics. The update allows developers to replace certain animal studies with validated alternatives, including in silico models, human cell assays, and organoid systems. This move builds on the FDA Modernization Act 2.0, which in 2022 removed the federal requirement for animal testing in some cases. The regulatory framework is now shifting toward more modern, mechanistic, and human-relevant tools in early-stage drug development.

According to Enke Bashllari, founder and managing director at Arkitekt Ventures, these shifts open up significant opportunities for startups focused on AI-driven prediction, modeling, and simulation in drug development.

These include PBPK/PD digital twins that simulate drug absorption, distribution, metabolism, and excretion to identify risks before human trials. AI toxicology, which applies deep learning to chemical structures and historical toxicology data, could be used to predict organ-specific safety concerns earlier in the pipeline. Virtual trial digital twins offer the potential to model synthetic patient populations using real-world data, genomics, and physiological parameters to improve trial design.

Other areas of opportunity include proteome profiling, where AI can map on- and off-target protein interactions across thousands of proteins to help prioritize compounds before wet-lab testing, and pharmacogenomics engines that integrate genomic, transcriptomic, and exposomic data to model response variability and support precision dosing. End-to-end predictive models that combine imaging, multi-omics, and clinical outcome data could also play a role in evaluating both safety and efficacy across the development pipeline.

Who is building, scaling, and using the infrastructure that could make this possible?

Let’s have a look at a crop of newer companies either generating massive biological datasets or building the tools to make sense of them. Think real-world evidence engines, privacy-first data exchanges, genomic stacks, AI discovery engines, and infrastructure for decentralized clinical trials.