AI in Biotechnology

Insilico Medicine's Rentosertib — the first antifibrotic small-molecule TNIK inhibitor, with both target and structure generated by generative AI — showed +98.4 mL mean change in forced vital capacity versus -20.3 mL on placebo at 12 weeks in the GENESIS-IPF Phase IIa trial (n=71), published in Nature Medicine. Patients on 60 mg QD not on concomitant antifibrotic therapy showed +187.8 mL. The first randomized clinical proof-of-concept that an AI-discovered target and AI-designed compound can deliver clinical efficacy in IPF.

Insilico Medicine's June 2026 white paper details how its Chemistry42 platform was used to design ISM2286, a first-in-class PKMYT1 PROTAC degrader published in Nature Communications. Generative models and kinase screening produced the warhead, refined by medicinal chemists, cutting off-target kinases from 46 to 11 versus RP-6306; AI-generated spirocyclic linkers raised solubility more than 100-fold.

Insitro combines automated laboratories generating multi-modal phenotypic cellular data with machine learning and generative AI to build phenotypic disease models, identify causal intervention points through human genetics, and deliver targeted therapies for the right patients.

Recursion operates BioHive-2, one of pharma's most powerful supercomputers built with NVIDIA, ingesting more than 50 petabytes of high-quality multimodal phenomics, transcriptomics, and proteomics data to power AI-enabled drug discovery across oncology and rare diseases.

Genentech and AWS built the gRED Research Agent on Anthropic's Claude 3.5 Sonnet in Amazon Bedrock Agents. It breaks scientific questions into multi-step workflows, uses RAG over PubMed and internal knowledge bases and queries internal APIs and databases, returning cited summaries. Work that took scientists weeks now takes minutes, and Genentech expects to automate over 43,000 hours of manual biomarker validation.

Shanghai Jincheng Technology's ManuDrive model, which predicts the next 180 hours of fermentation parameter curves with 99.9% accuracy, took over a 500-tonne fermentation tank at Chuanning Biotechnology in Xinjiang, raising output about 5% and cutting production volatility by 50%.

Amgen was an early adopter of Microsoft Copilot: what began as a small group of employees experimenting with it grew to 20,000 employees with access, used to cut repetitive office and lab tasks. In research, Amgen uses NVIDIA technology to build generative biology models and installed an NVIDIA DGX SuperPOD named Freyja at deCODE genetics in Reykjavik to analyse 200 petabytes of deidentified human data.

Insilico Medicine is pursuing 'pharmaceutical superintelligence' - using AI to transform the drug discovery process. The company recently launched its MMAI Gym that aims to train generalist large language models like ChatGPT and Gemini to perform as well as specialist models for drug discovery tasks.