AI drug discovery

Otsuka Holdings' Integrated Report 2026 says technology validation has started to expand the Cysteinomix drug discovery platform using generative AI. In the same report, Taiho Pharmaceutical's Tsukuba discovery division says it is working to introduce robotics and AI in the search for drug targets and compound design while expanding Cysteinomix.

Otsuka Holdings' Integrated Report 2026 states that Astex is using a proprietary dataset of more than 19,000 protein-ligand complexes to develop computational co-folding capabilities called AstexFold and generative AI models for fragment-based and structure-based drug design. The report also mentions the Pyramid platform integrating AI/ML methods. This is company self-reported, and the report does not state a deployment stage.

Novartis says that with generative approaches its teams can propose new molecular structures informed by protein information and prior chemistry and optimize multiple properties in parallel.

BMS says agents automate target identification as well as validation, saving weeks of manual work for its scientists.

Under its Predict First approach, BMS says AI-generated predictions inform the design of every small-molecule program and most large-molecule programs before bench work.

Daiichi Sankyo says it has introduced advanced AI models, including OpenFold-3, into its internal environment and provides them to researchers as cloud applications, adjusting models with the research organization so AI can be used in actual drug-discovery work.

Daiichi Sankyo says it has introduced AI agents into bioinformatics work and is using them to make exploration and analysis of life-science data more efficient and more advanced, while checking their usefulness on actual analysis work.

A research capability combining data, AI and machine learning, automation and robotics to speed discovery, presented at Moderna's 2026 Science Day.

Celltrion said it set up a dedicated AI-based new-drug organization the previous year and is applying AI in stages, with bioinformatics, to finding, validating and optimizing new-drug target candidates. The release also says the company is pursuing open innovation with external AI specialist firms, which it does not name.

J&J uses AI to screen chemical and biologic candidates and says this has halved lead optimization time and sped up two compounds.

BeOne's 2025 sustainability report says the company set budgetary targets to advance non-animal research tools through its in-house computational drug design and AI drug discovery groups. It says those groups now play an active role in every BeOne research program.

At its March 31, 2026 R&D Day, Astellas said NVIDIA's AI-powered supercomputing for small-molecule drug design has already been implemented.

In the same remarks, Astellas said gene therapy guided by AI will lead to precise organ targeting, reduced toxicity and enhanced treatment.

Hengrui's 2025 annual report says the company vigorously develops AI-driven drug discovery, which it abbreviates AIDD, and that AI technology is deeply integrated throughout its R&D pipeline.

Innovent's 2025 ESG report describes an AI+NGS wet-dry closed-loop antibody discovery platform that has been applied to multiple internal projects, shortening the antibody screening cycle and improving the efficiency of discovering high-quality candidate molecules.

Innovent says it has developed an AI-driven de novo design closed-loop system dedicated to small protein drug development. The same report says the system enables intelligent iteration from target elucidation to molecular optimization and has demonstrated a magnitude-level improvement in hit rates and R&D iteration speed in internal projects.

Merck KGaA's 2025 annual report says the company also uses AIDDISON, its generative-AI discovery platform, internally in the Healthcare business for early-stage drug discovery.

Roche and Genentech (gRED) run a Lab-in-the-Loop approach that connects biology and chemistry experiments with Roche's AI models, now enhanced with NVIDIA BioNeMo.

Partners: NVIDIA

AI and machine-learning algorithms inside Moderna's design and ordering platform that optimize mRNA sequences for manufacturability and quality.

Genmab says it has expanded its scientific focus to use data science and AI to aid in the discovery of new targets and biomarkers and to bolster its precision medicine and translational laboratory capabilities.

AI-driven target discovery engines in R&D. Sanofi says they produced seven new drug targets in a single year.

At its 2025 R&D day, Hengrui said AI-driven drug R&D has become an important innovation engine. It said it is relying on an AI target-discovery platform, an AI large-molecule design platform, an AI small-molecule generation and optimization platform, a specialized drug-research large model and AI drug-research agents to improve the efficiency of target discovery, molecule design and developability optimization.

On its risk-management page, Shionogi lists business-model and operations reform through practicing AI drug discovery among the responses to its digital-transformation risk.

The PharmaEssentia Innovation Research Center site says its data science team uses an extensive database and employs AI and ML to select checkpoint-receptor targets. The same page says the center uses AI and machine learning to establish databases so it can select novel targets and patients more wisely and screen molecules faster.

Astellas says it has been integrating AI into drug development since 2019 and that its Human-in-the-Loop platform, which combines human expertise, AI and robotics, is widely used in small- and medium-molecule discovery. The article says that by leveraging AI, employees have cut the time to refine a hit compound into a drug candidate by approximately 70% compared with traditional methods.

Novo combined AI with high-throughput experiments to assess one billion virtual molecules. The work led to a selective amylin compound.

Daiichi Sankyo said on January 21, 2025 that it had established a San Diego laboratory dedicated to robotics, automation and software for drug discovery. The release says the laboratory will collect data so discovery work can draw on AI, and so scientists can use analytical tools that are based on AI.

At the same AI Day, BioNTech unveiled novel AI Bayesian Flow Network models for protein sequence generation. The release describes the update as the launch of a novel BFN generative model.

InstaDeep's chief executive, speaking as part of BioNTech's AI Day, said a key focus is the DeepChain multiomics design platform. He said DeepChain is now open for external partnerships after successful application to several projects, including the mRNA-encoded antibody RiboMab platform.

BioNTech's AI Day release says the company is highlighting progress in deploying AI across its immunotherapy pipeline, including immunohistochemistry, DNA and RNA sequencing, proteomics, protein design and laboratory functional validation.

AbbVie says large language models let researchers engineer drug candidates in silico rather than relying solely on high-throughput screening.

AWS blog co-authored by Gilead's CIO says Gilead uses specialized LLMs to query and summarize literature and databases for drug-target assessment, aiming to cut assessment time by months.

Partners: AWS

Integrated AI algorithms read tumor and blood sequencing data to predict up to 34 neoantigens for each patient's individualized cancer therapy.

Merck KGaA launched AIDDISON, drug discovery software that it says combines generative AI, machine learning and computer-aided drug design to speed up drug development.

PharmaEssentia USA, announcing abstracts for the MPN Congress and ASH, said the program included an AI-based discovery analysis and listed a presentation described as a causal-AI dissection of RNA-seq data sets that pinpoints connections between myeloproliferative neoplasms and neurodegenerative diseases.

In the same release, Insmed says it is advancing an early-stage research engine that includes artificial intelligence-driven protein engineering, alongside gene therapy and protein manufacturing.

PharmaEssentia announced the PharmaEssentia Innovation Research Center in the Boston area and said data scientists at the center will use artificial intelligence and machine learning to inform new target and indication selections.

Roivant's prospectus says the in silico small-molecule discovery engine at Roivant Discovery is powered by QUAISAR, which it names as QUantum, AI and Structure-Activity Relationships. The same section says those capabilities predict how molecules interact and power in silico assays, and that the engine includes a suite of degrader-specific machine-learning tools the company has developed.