14 Biotechs Utilizing AI-based Research Platforms
According to our internal database, more than 500 life sciences companies in the categories "startups" and "scaleups" are actively using machine learning and deep learning-based predictive and generative capabilities to augment their research strategies in the life sciences, in particular, in drug discovery.
Some companies are incorporating AI tools to improve specific stages of drug discovery and development, for instance, to better model disease mechanisms or identify drug repurposing opportunities by analyzing omics data. Other companies are building entire end-to-end platforms for hypothesis generation, target discovery, de novo drug design, and even predicting clinical trial outcomes for newly designed candidates.
Earlier, we have addressed the question of what we consider to be a modern AI-driven drug discovery paradigm, and we also reviewed the progress over a decade of AI implementation by various drug discovery companies, concluding that the area is still in its early stages, with numerous setbacks and unrealistic expectations, but also room for future gains. Despite setbacks, companies keep iterating.
Below is a list of several noteworthy companies, which claim to be enabled by AI platforms and relying on computational capabilities to speed up, and possibly improve their therapeutics discovery work. In this article we mostly focus on clinical-stage AI drug discovery companies, many of which are also public companies.
Disclaimer: This post is not investment advice of any kind; it is for informational purposes only. Neither BPT Analytics Ltd. nor the authors of this post endorse or recommend any particular stocks or investment opportunities. Consult relevant professional service providers for any investment decisions at all times.
Absci (ABSI)
Absci is a U.S.-based biotech headquartered in Vancouver, Washington that applies generative models and an integrated wet lab to design therapeutic antibodies de novo. The company’s approach conditions models on antigen structure and epitope choice, then proposes CDR sequences “zero-shot,” followed by high-throughput experimental validation.
Its platform combines sequence- and structure-aware AI models with multiplexed build–measure loops, alongside a “reverse immunology” workflow that reconstructs antibodies from diseased tissues to infer targets and seed sequences for immuno-oncology. Model proposals are conditioned on epitope and antigen context, then validated via epitope-specific, high-throughput wet-lab screens.
In 2025 Absci became clinical-stage: ABS-101, an anti-TL1A antibody for inflammatory bowel disease designed with its generative workflow, began a randomized Phase 1 trial in healthy volunteers in May 2025. The company’s ABS-201 (anti-PRLR) program for androgenetic alopecia advanced with non-human-primate data and an accelerated plan to start a Phase 1/2a trial in December 2025. Another preclinical immuno-oncology program (ABS-301) originated via the reverse-immunology pipeline.
Current partner work spans AstraZeneca on oncology antibody design under a deal valued up to $247 million, Almirall on dermatology programs expanded in August 2025 to a second target following delivery of AI-generated de novo antibodies, Twist Bioscience on joint AI-enabled antibody discovery workflows, and AMD with Oracle on an OCI-based compute collaboration to scale AI model training and molecular simulation workloads.
AbCellera (NASDAQ: ABCL)
AbCellera is a U.S./Canada‑based biotech company (headquartered in Vancouver) that uses AI, single‑cell analysis and deep sequencing to discover therapeutic antibodies. It went public in December 2020 via an IPO that raised approximately US$555.5 million.
Its platform screens millions of single B cells using microfluidics and high-throughput assays to measure secreted antibody function in real time. Paired heavy and light chain (VH/VL) sequences are recovered from each cell, preserving natural diversity. This enables rapid identification of antibodies with specific functional properties, such as affinity, specificity, cross-reactivity, and blocking activity, against complex targets including GPCRs, ion channels, and other multipass membrane proteins. Computational tools then model and optimize lead candidates for developability, manufacturability, and potency, accelerating the path from immune response to preclinical validation.
In 2025 the company advanced internal programs such as ABCL635 (NK3R) and ABCL575 (OX40L) into Phase 1 clinical trials.
BioAge Labs (NASDAQ: BIOA)
BioAge Labs is a clinical-stage biotech using machine learning on longitudinal human multi-omics and health records to find targets for metabolic aging (obesity, cardiometabolic risk). The company went public in September 2024 with gross proceeds ~$198M and now reports a pipeline centered on a CNS-penetrant NLRP3 inhibitor and next-gen APJ agonists.
Recently, BioAge began a randomized, placebo-controlled Phase 1 study in healthy volunteers of BGE-102—an oral, brain-penetrant NLRP3 inhibitor aimed at obesity. BioAge also broadened its APJ program—securing an option with JiKang to take a ≥10×-over-apelin APJ nanobody through IND-enabling studies.
BioAge Labs is building a data-first platform for metabolic aging. The company follows healthy people over time, coupling longitudinal multi-omics (transcriptomics, proteomics, metabolomics, etc.) with repeated measures of physical, cognitive, and functional performance to map “healthspan trajectories” as they shift toward disease. AI/ML models mine these datasets to surface drug targets across known and novel aging pathways; hits are validated in-house using translational studies in naturally aged mice, then advanced with a tech-forward discovery stack.
In June 2025, BioAge expanded its platform via a collaboration with Age Labs to deeply profile >17,000 samples from 6,000+ HUNT Biobank participants—data to which BioAge has exclusive drug-discovery rights—adding millions of molecular measurements that track the transition from health to cardiometabolic disease over decades.
In December 2024, BioAge Labs signed a multi-year deal with Novartis to use BioAge’s longitudinal aging datasets (plus Novartis’ exercise biology expertise) to find and validate drug targets for aging-related diseases, valued at over $500 million. Other collaborations include Lilly ExploR&D agreement for antibody development in metabolic aging.
Generate:Biomedicines
Generate:Biomedicines is a privately held biotech that was founded in 2018 by Flagship Pioneering, and is headquartered in Cambridge, Massachusetts. The company focuses on applying generative‑AI and biological engineering to create novel therapeutic proteins — effectively shifting from traditional drug discovery toward “programmable” biologics.
The heart of their technology is the Generate Platform, which integrates machine learning, high‑throughput molecular build/measure loops, and structural determination at scale. The platform works as a continuous cycle: (1) Generate – algorithms design protein sequences de novo or optimize existing scaffolds; (2) Build – these sequences are synthesized and expressed; (3) Measure – functional and biophysical assays quantify binding, specificity, immunogenicity, manufacturability; (4) Learn – data feed back into models to improve future design.
As per company claims, their approach enables access to traditionally hard‑to‑drug modalities (e.g., novel epitopes, complex protein interfaces, non‑natural scaffolds), and they aim to optimize not just potency but also developability, dosing interval, immunogenicity, and manufacturability from the start.
On the pipeline side, Generate has advanced several programs into the clinic: for example, GB‑0895 (an anti‑TSLP antibody for asthma) and GB‑7624 (anti‑IL‑13 for atopic dermatitis) are in or entering Phase 1. They also announced GB‑0669, a monoclonal antibody targeting a conserved SARS‑CoV‑2 spike region, which has entered Phase 1 proof‑of‑concept.
Insilico Medicine
Founded in 2014, Insilico Medicine (Private; HKEX listing application filed May 2025) is a Cambridge, MA–based clinical-stage biotech using an end-to-end AI stack for target discovery, small-molecule design, and clinical forecasting.
Insilico’s Pharma.AI suite combines: PandaOmics for multimodal target/biomarker discovery; Chemistry42 for generative small-molecule design and physics-based optimization; and InClinico for trial-outcome prediction and protocol sensitivity analyses. In 2025 the company also released Nach01, a chemistry-aware foundation model on AWS Marketplace, and showcased an automated lab program that includes a bipedal “AI scientist” for data generation. These tools support partnered discovery and Insilico’s internal pipeline.
Insilico reported 22 preclinical candidates nominated in 2021–2024 with an average ~13 months to PCC (vs. 2.5–4 years traditionally), 10 programs that cleared IND to enter human trials, and an ENPP1 inhibitor (ISM5939) that moved from design to IND in ~3 months; its lead, rentosertib for idiopathic pulmonary fibrosis, showed in Phase IIa data published in June 2025 in Nature Medicine a manageable safety profile; and in August 2025 the company said it had completed IND-enabling work for ISM8969, an oral, brain-penetrant NLRP3 inhibitor for Parkinson’s disease. The company also recently unveiled eight AI-designed oral cardiometabolic programs.
In January 2025, Insilico Medicine licensed a second AI-designed oncology candidate to Menarini’s Stemline Therapeutics for $20 million upfront and up to $550+ million in milestones; the preclinical asset reportedly shows broad activity across solid tumors and follows MEN2312, a KAT6 inhibitor designed using the same AI platform that entered Phase 1 in 2024 for metastatic ER+/HER2- breast cancer.
Recently, Insilico raised $110 million in Series E and unveiled “Supervisor,” a humanoid lab robot designed to take over routine bench tasks, and entered a research and licensing collaboration with Eli Lilly valued at over $100 million in potential payments, combining Insilico’s Pharma.AI platform with Lilly’s disease and development expertise to jointly advance new therapeutic programs.
Lantern Pharma (NASDAQ: LTRN)
US-based pharmaceutical company Lantern Pharma specializes in developing precision oncology therapies leveraging artificial intelligence and genomics. Founded in 2013, Lantern Pharma closed a $26 million IPO in 2020.
Lantern Pharma has two directions in its drug discovery and development pipeline: developing new classes of precision cancer drugs with novel mechanisms of action and “recycling” previously unsuccessful cancer drugs using machine learning algorithms, genomic data, and novel precision oncology biomarkers.
Read also: How Lantern Pharma's AI-driven Approach Enables Precision Oncology
Lantern Pharma developed the RADR (Response Algorithm for Drug Positioning and Rescue) platform, which is an integrated data analytics, experimental biology, biotechnology, and machine-learning-based platform. It is a sophisticated tool in assisting with defining and developing combination strategies among drugs in development and those that are approved for a range of oncology indications—now leveraging more than 200 billion oncology-specific data points, up from 100 billion reported earlier in 2025.
Earlier this year, Lantern also released predictBBB.ai, an open-access BBB-permeability predictor, expanding RADR’s scope beyond response modeling.
As of now, Lantern moved three AI-guided cancer programs forward: LP-184 wrapped its Phase 1a enrollment is lining up next-stage trials in hard-to-treat solid tumors like triple-negative breast and bladder cancer; LP-284 for non-Hodgkin lymphoma saw a first complete metabolic response in a heavily pretreated patient and shared updated results this fall; and LP-300 for never-smoker lung cancer kept enrolling, expanded into East Asia with enrollment completed in Japan, and continued to show early signs of benefit in the initial cohort. One participant—a 70-year-old never-smoker with advanced NSCLC—achieved a complete response in lung and adrenal lesions that has persisted for nearly two years across 21 treatment cycles, with no dose-limiting toxicities or significant adverse events reported.
Moderna Inc (NASDAQ: MRNA)
Moderna is a biotech company founded in 2010 focusing on computationally-driven drug discovery and vaccine development based on messenger RNA technology. In 2018, Moderna closed a record-setting IPO, raising a total of $604.3 million.
Their research engine combines a proprietary artificial intelligence-driven digital engine for drug discovery and vaccine development, integrated into a highly automated production facility to develop mRNA vaccines and therapeutics. Using Moderna’s technology, scientists can select a protein in the human proteome to be engineered or design novel proteins.
A web-based mRNA Design Studio enables rapid design of multiple mRNAs, starting from the protein sequence. Additionally, their mRNA Early Development Engine services combine internal and external capabilities for process development, toxicology studies, global regulatory interactions, and clinical study preparation and execution for mRNA vaccines and therapeutics.
The company became famous in 2020 when it managed to develop and commercialize the COVID-19 vaccine, which gained global recognition. Since then, Moderna added an FDA-approved RSV shot, mRESVIA (mRNA-1345)—first cleared for adults ≥60 on May 31, 2024, and later expanded on June 12, 2025 to at-risk adults 18–59.
Advancing programs built on its AI-enabled mRNA Design Studio/early-development engine include the seasonal flu vaccine mRNA-1010 and the individualized cancer vaccine V940/mRNA-4157, now in Phase 3 in adjuvant melanoma (INTerpath-001) and NSCLC (INTerpath-009) with Merck.
Neumora Therapeutics (NASDAQ: NMRA)
In 2023, Neumora Therapeutics, a clinical-stage biopharmaceutical company focused on developing treatments for neuropsychiatric disorders and neurodegenerative diseases, announced the pricing of its initial public offering (IPO) at $17.00 per share, raising approximately $250 million.
Neumora describes a precision-neuroscience platform that applies proprietary machine-learning methods to large, multimodal patient datasets like genetics, imaging, EEG, digital and clinical records to derive “Data Biopsy Signatures” that map disease biology and link them to “Precision Phenotypes” used for patient stratification and trial design. The company’s 10-K details a “Precision Toolbox” built on ~1 petabyte of longitudinal, multimodal patient data intended to identify biomarkers, enroll enriched cohorts, and optimize study designs. The platform also ingests human-genetics insights via Amgen’s deCODE collaboration to inform targets and cohort definition.
Neumora's approach integrates translational, clinical, and computational methods to advance precision medicine in neuroscience, with a pipeline of seven clinical and preclinical programs targeting novel mechanisms for a range of brain diseases, such as schizophrenia, Parkinson’s and Alzheimer’s disease, as well as several candidates also having potential in ASL and obesity.
Recently, Neumora reported preclinical results for NMRA-215, an oral, brain-penetrant NLRP3 inhibitor, showing up to ~19% weight loss as monotherapy in diet-induced obese mice and up to ~26% when combined with semaglutide. The company attributes the effect to central target engagement and positions NMRA-215 as a potential add-on to incretin therapy or a lower-dose alternative. These are mouse data—clinical testing is planned to start in 1Q 2026.
OWKIN
Owkin is a Franco-American techbio using federated and multimodal AI to discover targets, optimize trials, and deploy digital pathology diagnostics. The company is privately held (no IPO as of Nov 10, 2025). Reported funding totals ~$304.1M with the latest $230M Series B raised across two separate closings of the round.
Owkin’s proprietary platform, branded as “Owkin K” and recently publicised via its “K Pro” agentic-AI tool, combines multimodal patient data, foundation models and reinforced learning to map complex biology across molecular, tissue and disease scales. The company says it ingests rich, research-grade patient data (including spatial omics, single-cell and pathology imaging) sourced from academic institutions and harmonised for AI training.
Once data is ingested, Owkin applies a layered AI stack: multimodal embeddings to link cells, tissues, patients and drugs; fine-tuned large language-models and reinforcement-learning agents to “uncover hidden causal relationships” and propose actionable insights; automated lab perturbation and real-world clinical data feed back into the system for continuous improvement. The platform supports three major domains: drug discovery (target & indication identification), clinical development (trial design, synthetic controls, cohort stratification) and diagnostics (digital-pathology tools for biomarker detection and outcome prediction). Owkin also helped build MELLODDY, a first-of-its-kind, industry-scale federated learning platform where 10 pharma companies trained shared drug-discovery models.
In November 2024, Owkin and Proscia agreed to integrate Owkin’s MSIntuit CRC v2 into Proscia’s Concentriq platform to pre-screen colorectal cancer patients for microsatellite instability, with a Research Use Only version rolling out in the U.S. to help labs scale MSI screening and cut unnecessary confirmatory tests.
Owkin’s current collaboration slate includes Sanofi with a $180 million equity investment with a multi-year oncology alliance announced in November 2021, expanded in 2024 to apply Owkin’s AI to drug positioning in immunology, as well as a June 2022 multi-year deal to use Owkin’s AI for clinical-trial optimization Bristol Myers Squibb collaboration, and a collaboration around the MSIntuit CRC digital-pathology test with MSD (Merck).
Recursion Pharmaceuticals Inc (NASDAQ: RXRX)
Founded in 2013, Recursion Pharmaceuticals is a Utah-based clinical stage "digital biotech," possessing a sophisticated robotized biology experimentation drug discovery platform utilizing machine learning technologies, including high-throughput imaging analytics and high-content screening data modeling. The biotech company went public in April 2021, having raised $502 million.
Recursion’s active internal programs now concentrate on oncology and rare disease. With the recent acquisition of Exscientia, the combined company references partnered portfolios with Sanofi, Merck KGaA, Roche/Genentech and Bayer.
Recursion’s “Operating System” is an end-to-end setup that turns wet-lab biology into a continuous data loop: robots run standardized experiments, the results are converted into rich digital readouts, and AI models, including purpose-built foundation models and LLMs acting as task “agents”, predict new targets, compounds, or follow-up assays and trigger the next lab runs. This closed loop lets the platform iterate from target finding to trial enrollment, with each cycle adding to the proprietary biology/chemistry datasets the company uses for training. The Recursion OS also integrates high-volume patient datasets from Tempus, Helix, and HealthVerity, supporting trial design and recruitment.
Recursion reports a ~65-petabyte proprietary dataset, which is used to train models that propose new targets and design optimized molecules, with processing handled on BioHive-2, a supercomputer built with NVIDIA.
In June 2025, Recursion and MIT launched Boltz-2, an open-source AI model that predicts a protein’s 3D shape and how strongly a drug candidate binds to it in a single pass. Trained on ~5 million binding measurements plus simulation data, it delivers free-energy–like accuracy at roughly 1000× faster speeds, and supports simple user guidance (e.g., contact or template hints). Additionally, Recursion, together with Roche and Genentech, has created the first “Microglia Map” — a massive AI-analyzed dataset that shows how more than 17,000 genes affect microglia, the brain’s immune cells that play a key role in diseases like Alzheimer’s and Parkinson’s.
Recursion set a leadership handover for January 1, 2026—Najat Khan will become CEO and President (continuing as a director), while co-founder Chris Gibson moves to Chairman and interim Executive Advisor; Rob Hershberg becomes Vice-Chairman and Lead Independent Director. The transition is framed as positioning the company to scale its OS platform and pipeline.
Relay Therapeutics Inc (NASDAQ: RLAY)
Founded in 2016, Relay Therapeutics is a US-based company that grew up specializing in developing an artificial intelligence-driven allosteric drug discovery platform intended to apply computational techniques to modeling protein motion. They went public in 2020 with a $460 million IPO.
Relay Therapeutics’ artificial intelligence-driven Dynamo platform centers on the idea that proteins are not static structures but constantly moving systems, and that understanding their motion can reveal new ways to design drugs. The process begins with building detailed experimental maps of how target proteins shift and change shape using tools such as cryo-electron microscopy and ambient-temperature X-ray crystallography. These experimental results feed into molecular dynamics simulations that model how full-length proteins move over time, helping pinpoint hidden or transient binding sites and generate hypotheses about how to best modulate a target’s function.
Once potential binding regions are identified, the platform integrates experimental screens and computational modeling to search for suitable chemical starting points. Machine learning models trained on this data, together with a proprietary DNA-encoded library system called REL-DEL, rapidly suggest molecules likely to interact effectively with the target. Dynamo then connects advanced ML models with physics-based simulations and medicinal chemistry workflows to refine these hits into viable candidates, optimizing for potency, selectivity, and drug-like properties.
In June 2024, Relay announced a collaboration with Pfizer to test its AI-designed cancer drug RLY-2608 together with Pfizer’s atirmociclib and the standard therapy fulvestrant in a type of advanced breast cancer driven by specific genetic mutations.
At ASCO 2025, Relay reported that patients taking RLY-2608 plus fulvestrant went a median ~10 months before their cancer worsened, with side effects described as generally manageable. The company said this supports starting a Phase 3 ReDiscover-2 trial and it is continuing tests of three-drug combinations that add atirmociclib or ribociclib.
Schrödinger Inc (NASDAQ: SDGR)
Headquartered in New York, Schrödinger is a company developing molecular design software for pharmaceutical, biotech, and materials research. Founded in 1990, the company went public in 2020 with a $232 million IPO.
Schrödinger builds drug and materials candidates with a physics-driven software stack that combines molecular modeling, simulation, and AI. The core engine, FEP+ (free energy perturbation), estimates how tightly a molecule will bind to a protein with near-experimental accuracy (around 1 kcal/mol), which helps rank ideas, pick better chemical cores, and tune properties like potency, selectivity, and solubility before going to the lab. Teams use it from early structure work through hit discovery and lead optimization, and for protein tasks such as antibody refinement, peptide binding, and enzyme design. Active-learning loops let the system triage very large libraries in silico, cutting down the amount of experimental screening needed.
Schrödinger has some fully-owned and collaborative drug discovery programs in a broad range of therapeutic areas. Schrödinger’s current pipeline spans blood cancers (MALT1 inhibitor SGR-1505), broad solid tumors (Wee1/Myt1 inhibitor SGR-3515), RAS-driven tumors (SOS1 program), EGFR C797S–mediated resistant lung cancer, inflammasome-driven diseases (NLRP3 inhibitor), and Parkinson’s disease (LRRK2), and undisclosed discovery programs.
At the 2025 European Hematology Association meeting, Schrödinger said its AI-designed MALT1 inhibitor SGR-1505 was generally safe and showed early signs of benefit as a single drug in several relapsed or refractory B-cell cancers, including chronic lymphocytic leukemia and Waldenström macroglobulinemia. The candidate also recently received FDA Fast Track designation.
In November 2024, Schrödinger and Novartis signed a multi-target discovery deal that brings Schrödinger $150 million upfront (plus up to $2.3 billion in milestones and tiered royalties) and expands Novartis’s three-year software license to run Schrödinger’s platform at scale. In July 2025, Ajax Therapeutics and Schrödinger expanded their 2019 partnership to add another Janus kinase (JAK) target with the first joint drug—AJ1-11095 for myelofibrosis—already in Phase 1.
XtalPi (HKEX: 2228.HK)
XtalPi, an AI drug discovery company based in Shenzhen, with its initial IPO filing, has commenced trading on the Hong Kong Stock Exchange on Thursday, June 13, 2024. Traded on HKEX as XtalPi Holdings Limited (stock short name XTALPI-P, code 2228) In preparation for the listing, XtalPi raised 896 million Hong Kong dollars (approximately $115 million), positioning it as the third-largest initial public offering (IPO) in Hong Kong that year.
Founded in 2015 by three quantum physicists from the Massachusetts Institute of Technology (MIT), XtalPi’s core platform, ID4 (Intelligent Digital Drug Discovery & Development), links physics-based molecular simulation with machine-learning models and autonomous “wet-lab” robotics in a closed loop. In practice, AI models propose chemical designs or select hits; quantum/first-principles calculations and high-throughput virtual screening rank them; and automated labs run synthesis and assays, sending results back to retrain the models (active learning). The stack is positioned to handle potency, ADME/PK, and solid-form risks early, and XtalPi runs the workflow at scale across cloud/HPC and multi-station robotics labs.
The platform is used in pharma collaborations (e.g., Eli Lilly) to deliver de novo candidates; notably, in November 2025, Eli Lilly signed another bispecific-antibody deal with Ailux (a XtalPi subsidiary), paying up to $345 million total. In August 2025, XtalPi finalized a deal with DoveTree Medicines worth up to $6 billion to use XtalPi’s AI-and-robotics discovery platform across oncology, immunology, neurology, metabolic and other programs.
XtalPi also highlights the platform’s extension into materials and other domains. In June 2025, XtalPi and Pfizer expanded their partnership to build a next-generation small-molecule modeling platform that blends physics-based methods with AI. The company markets this as an end-to-end service: design, make, test, learn—under one loop, with autonomous labs in Shenzhen/Shanghai (and build-outs in Shanghai and Cambridge, MA).
In September 2025, XtalPi and PharmaEngine won approvals in Taiwan and Australia to start a Phase 1 study of PEP08, an AI-designed second-generation PRMT5 inhibitor for advanced solid tumors.
The company raised a total of $784 million since its inception. Its latest private fundraising was a $400 million Series D round in October 2021, valuing Xtalpi at approximately $2 billion.
Tempus AI (NASDAQ: TEM)
Tempus AI, Inc., a technology company focused on advancing precision medicine through AI, announced its IPO on June 14, 2024, raising $410.7 million.
Founded in 2015 by Eric Lefkofsky, a co-founder of Groupon, Tempus AI was initially focused on oncology. Later, it expanded its platform into neuropsychology, radiology, and cardiology, to establish a comprehensive presence in all major disease areas worldwide. This expansion is fueled by the company's data acquisition capabilities, strategic partnerships with healthcare providers and industry associations, and a number of funding rounds totaling over $1.3 billion.
Tempus’ AI suite is a connected set of in-workflow tools spanning care and research: Tempus One (a generative-AI assistant) lets clinicians and researchers query structured and unstructured records and build task-specific agents via Agent Builder, while integrating directly with electronic health records and surfacing guideline context in-EHR. Lens provides data exploration and no-code analytics on Tempus’ multimodal datasets; Pixel automates lesion segmentation, tracking, and reporting from medical images; Next flags care-pathway gaps and notifies teams inside clinical workflows; Link uses AI to scan chart data and speed trial matching; Hub and Now handle test ordering, results, and EHR delivery; and assay/algorithm offerings (e.g., xT, xF, IPS, HRD, PurIST) supply the genomic inputs that feed these applications.
In September 2025, Tempus reported real-world validation of PurIST, an RNA-based test that classifies advanced pancreatic cancers into “classical” or “basal” subtypes to guide first-line chemotherapy.
Recently, Tempus launched “Loop,” an oncology-focused platform that uses multimodal real-world patient data to cluster subpopulations, applies AI/systems-biology models to generate target hypotheses, and then validates them in patient-derived organoids via high-throughput CRISPR screens. The company also announced a separate collaboration with AstraZeneca and Pathos to build a large multimodal oncology foundation model trained on de-identified Tempus data, with $200 million in combined fees for data licensing and model development.
A Reality Check for AI in Drug Discovery
While the promise of AI in drug discovery is significant, the reality is that successful implementation and the ability to achieve practical improvements vary dramatically from one use case to another and require quality data, time, highly skilled talent, and a lot of experimentation and validation. Even leading players in this segment are not immune to setbacks. In particular, several drug candidates, advertised as AI-generated, did not meet clinical trial endpoints, and their development was aborted.
Also, with the exception of several companies like Schrödinger, Recursion Pharmaceuticals, and Moderna, the rest of publicly traded AI in drug discovery stocks are showing losses since their IPOs.
That being said, the sector is experiencing consolidation, while companies like Insilico Medicine are demonstrating substantial progress with their AI-generated pipelines.
What is clear, though, is that the advent of AI in drug discovery and development is not a story of a revolution but rather an evolution. Evolution takes time, and it is iterative.
Topic: AI in Bio
