Top market ‘AI in biotech’ stories
AI catches errors in scientific papers, untangles gene regulation, and promises to upend drug discovery via a generalist model
We’re introducing a new monthly series from Ardigen—your curated selection of the latest news in AI for biotech and pharma. In just 3–4 minutes, catch up on the innovations reshaping drug discovery and biomedical research. We’ll publish fresh insights on the second Tuesday of each month, so you can stay informed and inspired. Happy reading!
In today’s edition:
- 🧬A global alliance for spatial transcriptomics
- 💻Borzoi: a unifying model of gene regulation
- 💊A drug discovery generalist model from NVIDIA
- 🤖AI is catching errors in research
GESTALT: A global alliance driving innovation in spatial biology
A new global initiative is set to accelerate progress in spatial transcriptomics. The Global Alliance for Spatial Technologies (GESTALT) has launched a central hub for the Spatial Tissue Profiling Community, bringing together over 1,000 members across 40 countries. This rapidly growing alliance aims to unite assay developers, bioinformaticians, computational biologists, clinicians, and industry leaders to foster collaboration, knowledge-sharing, and technological advancements.
Introducing Borzoi: A breakthrough in gene regulation modeling
Calico Life Sciences has introduced Borzoi, a neural network model that predicts RNA-seq coverage directly from DNA sequences, integrating transcription, splicing, and polyadenylation in a unified framework. By leveraging deep learning on large datasets like GTEx and ENCODE, Borzoi outperforms existing models in its ability to capture cell type- and tissue-specific patterns, making it a powerful tool for advancing our understanding of gene regulation across species.
GenMol is a new generalist foundation model by NVIDIA that aims to advance AI-driven molecular generation in drug discovery. GenMol introduces a versatile molecular generative framework using discrete diffusion and non-autoregressive decoding to enhance flexibility and efficiency in drug discovery tasks. It outperforms prior models in de novo generation, fragment-constrained design, and lead optimization, offering a unified approach for various stages of molecular design.
AI-powered error detection aims to improve scientific integrity and prevent misinformation
Last year, a study falsely claimed black plastic utensils contained dangerous levels of cancer-linked chemicals. To combat similar types of misleading findings, AI-powered tools are being developed to detect errors in scientific research. The Black Spatula Project and YesNoError use large language models (LLMs) to analyze thousands of studies for issues in calculations, methodology, and referencing. Supporters believe AI could improve research integrity by catching errors before publication, but critics warn of unintended consequences, such as risk of politicized scrutiny.
