Rewriting Drug Discovery with an AI-Multi-Omics Approach

Episode Summary

Executive Summary: The episode features Sapient CSO Mo Jane explaining why genomics alone cannot explain most disease and how Sapient uses AI-driven multi-omics, especially mass spectrometry-based proteomics, to identify better targets, biomarkers, and patient subgroups across drug discovery, translation, and clinical development. The conversation argues that richer biological data plus AI can reduce failure, accelerate development, and improve precision medicine.

Main Topics: Why genomics is insufficient for common disease (Priority: 5/5): Jane argues that genetics explains only a minority of disease risk for most common illnesses, so a broader multi-omics view is needed to capture the dynamic biology that drives disease. Sapient’s origin and platform (Priority: 5/5): He traces Sapient to his clinical and academic work, where large-scale mass spectrometry revealed predictive biological signals that could be commercialized for pharma and biotech. Multi-omics across the drug-development pipeline (Priority: 5/5): The discussion explains how Sapient applies proteomics and related measurements from target discovery through preclinical testing, translational work, clinical trials, and companion diagnostics. AI as an enabler, not a standalone solution (Priority: 4/5): Jane says AI currently helps most with chemical optimization, but its larger value comes when paired with high-quality multi-omic data to solve target ID, validation, translation, and patient selection. Better target discovery and novel biology (Priority: 5/5): Sapient claims deeper proteomic measurement can reveal more tumor-specific targets, including isoforms and post-translationally modified proteoforms not seen with conventional methods. Bridging discovery to clinic (Priority: 4/5): The company uses a service model plus internal data assets and assay development to help clients move from exploratory findings to regulated clinical tests. Precision medicine and future impact (Priority: 4/5): Jane forecasts that the near-term impact will be better diagnosis, patient stratification, and matching therapies to disease biology, with potential to lower development costs over time.

Key Arguments: Genomics explains only a minority of disease risk for common diseases, leaving most biological causation unresolved. Disease is better understood as a dynamic, multi-layered process involving proteins, metabolites, lipids, and cell-cell signaling, not just DNA variation. Mass spectrometry can measure thousands of molecules in complex specimens and generate predictive signatures at scale. Drug development fails because end-stage pathology is not the same as the biological pathway that caused the disease; therapies must target mechanisms, not just diagnoses. Multi-omics can improve target identification, assay development, preclinical readouts, patient enrichment, and companion diagnostics. AI is useful, but its current impact is limited when it is only applied to chemical screening rather than the whole development pipeline. Sapient differentiates itself by combining deep proteomic measurement, internal clinical data, and statistical/AI analysis to translate discoveries into regulated assays and clinical utility. The biggest opportunity is not merely making targets druggable, but identifying the right target and the right patient for each therapy.

Data Points: Population-attributable disease risk explained by genomics: 15-20% - Jane says genomics explains only a minority fraction of risk for common diseases such as heart disease, cancer, neurodegeneration, and autoimmune disease. Unexplained disease risk: 80-90% - He uses this as the gap that multi-omics aims to fill beyond genomics. Drug attrition rate in clinical trials: ~90% - He says about 90% of therapeutics entering clinical trials ultimately fail to gain FDA approval. Tumor proteins measured in a typical specimen: 12,000-13,000 proteins - Example of the depth of proteomic profiling in tumor versus normal comparisons. Proteins altered in typical tumors: 2,000-3,000 proteins - He describes the scale of differential expression observed between tumor and matched normal tissue. Known and emerging oncology drug targets: Only a couple hundred - He contrasts the limited set of current targets with the much larger proteomic landscape. Proteome coverage in complex tissues: 14,000-16,000 proteins - Sapient claims its mass spectrometry platforms can reach this scale in complex tissues. Scale of early datasets: 1,000 to 10,000 samples - He cites the early ability to generate and analyze large datasets at scale. Oncology driver mutation prevalence: 15-20% of tumors - He notes that only a minority of tumors have a clear single driver mutation. AI horizon for acceleration: 2-3 years - Jane predicts notable acceleration in AI-derived therapeutics over this timeframe.

Pivotal Quotes: "Your zip code remains a better predictor of how healthy you'll be over the course of your life than your genetic code does." — Mo Jane: Used to argue that environment and other non-genetic factors explain much more disease risk than genomics alone. "It's like trying to solve a puzzle when you only have 10-20% of the pieces and you don't know what the final puzzle image even looks like." — Mo Jane: His analogy for why genomics alone cannot solve common disease biology. "The real question is no longer: what's the best drug modality, or can I actually drug this target? The question really now is, what's the best target? And then, what's the best patient?" — Mo Jane: Summarizes the company’s view that target selection and patient selection are now the core bottlenecks.

Implications: For pharma and biotech, the message is to move beyond single-gene thinking and use multi-omics plus AI to improve target selection, biomarker development, and patient stratification. The likely payoff is faster, more successful, and more personalized drug development.