Astellas needed to shorten early-stage drug discovery by predicting antibody properties and generating candidate molecular structures faster, but generic protein language models and conventional conformational screening methods were too inaccurate or too slow for antibody-specific research.
Reported impact
Astellas improved antibody-property prediction accuracy, trained heavy- and light-chain models in about 65 and 37 hours, and increased molecular conformational screening speed by 50-60x, reducing jobs that once took hours to a day down to as little as 15 seconds.
Astellas built an antibody-focused generative AI workflow on NVIDIA BioNeMo to support a broader human-in-the-loop drug-discovery platform. The company created its own antibody language model, astABpLM, by using BioNeMo's ESM-1nv model as the starting point, preprocessing 2.4 billion antibody sequences from the OAS database with NVIDIA RAPIDS, and training separate heavy-chain and light-chain models on an NVIDIA DGX A100 system. In parallel, the company used generative AI with torsional diffusion to create diverse 3D compound structures for low- and mid-molecular-weight molecules, including PROTACs, and is extending the work through the Tokyo-1 innovation hub's DGX H100 environment. This architecture lets Astellas combine sequence modeling, molecular-structure generation, and robotics-ready drug-discovery workflows without spending research time on low-level algorithm or parameter tuning.
Antibody sequences from OAS are preprocessed with RAPIDS, BioNeMo's ESM-1nv is trained into astABpLM on DGX A100 systems, the resulting model predicts antibody properties for candidate selection, and separate generative models produce 3D compound structures to feed the wider human-in-the-loop drug-discovery process.
"We now get results in as little as 15 seconds, compared to the previous environment, which took several hours to a day."
Astellas Pharma Inc. has developed its own antibody language model, astABpLM, utilizing NVIDIA's generative AI framework for drug discovery, BioNeMo, to efficiently predict the properties of new antibodies.
One NVIDIA DGX A100 is used as the hardware. The heavy chain model astABpLM_VH and light chain model astABpLM_VL completed training in approximately 65 and 37 hours, respectively.
The company's proprietary workflow for molecular conformational screening of compounds has enabled a 50- to 60-fold increase in speed compared to conventional methods.
The data size was 2.4 billion sequences, which were preprocessed using NVIDIA's RAPIDS™ suite for data science to prepare the dataset for training.
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
NVIDIA Sample Sep 22, 2026 DELFI Diagnostics DELFI Diagnostics needed to speed genomic analysis for cancer screening and therapeutic monitoring, but CPU-bound alignment and somatic variant-calling steps created turnaround-time and cost bottlenecks that limited how… 10–20 business days DeepSomatic · MutectCaller · NVIDIA Parabricks NVIDIA Sample Jul 10, 2026 Astellas Pharma Inc. Astellas needed to shorten the slow handoff between experimental scientists and computational chemists in early drug discovery so more researchers could use structure prediction and affinity screening without waiting on… More than 50 regular users and over 2,000 total users within four months NVIDIA BioNeMo · NVIDIA Boltz-2 NIM · NVIDIA DGX H100/H200 NVIDIA Sample Apr 23, 2026 Biofy Hospitals needed pathogen identification and antibiotic-resistance analysis much faster than five-day culture and antibiogram workflows, but CPU-bound genomic processing and limited AI expertise made it hard to turn… 100,000+ synthetic DNA sequences generated by the biology model Abby · H100 Tensor Core GPUs · NVIDIA CUDA-accelerated libraries 