Company news
Outer Biosciences is pairing living human skin with an AI model
The startup wants repeated experiments on viable human skin to improve predictions about how compounds will behave before clinical testing.
Independent reporting2 sources
Topic
Biological foundation models
General-purpose learned representations for molecules, cells, tissues, and other biological systems.
News
Recent clusters
Company news
Network Bio launches with $50 million and an NVIDIA model collaboration
Network Bio paired a $50 million launch financing with plans to build a cell-free RNA foundation model alongside NVIDIA.
Independent reporting4 sources
Company news
GenScript says its AI drug-discovery business doubled in the first half
GenScript reported that its AI-enabled drug-discovery business doubled year over year, without disclosing standalone segment revenue.
Reported development1 source
Policy & regulation
EU research centre maps AI-for-biology capabilities and policy gaps
A Joint Research Centre report calls for realistic capability claims, transparent reporting, and reproducible validation of biology AI systems.
Reported development1 source
Research
Related papers
Finetuning Foundation Models for Temporal Clinical Transcriptomics Data
Mathur and colleagues fine-tuned foundation-model gene embeddings and used temporal graph neural networks to distinguish treatment-response patterns across four inflammatory diseases.
Deciphering the comprehensive relationship between 5′ UTR and 3′ UTR sequences with deep learning
Suga and colleagues combined pretrained RNA-language-model representations with contrastive learning to identify coordinated relationships between 5′ and 3′ untranslated regions.
scDisent: regulatory-aware disentangled representation learning for multi-omic single-cell analysis
Xi, G. reports that scDisent separates expression-linked and regulation-linked variables instead of compressing multi-omic measurements into one latent space.
Robustness to nuisance perturbations enables unsupervised evaluation of single-cell foundation models
Sallam, A. and colleagues report that an unsupervised stress test separated single-cell foundation models that looked nearly equivalent on standard downstream benchmarks.
A novel benchmark dataset for enzyme function prediction reveals the limitations of state-of-the-art models
Joao Sartori and colleagues report that enzymARC tests whether enzyme-function predictors reject structure-guided decoys whose catalytic machinery has been disrupted.
GRASSP: RNA Language Model–Enhanced Graph Attention with Adaptive Gating for RNA–Small Molecule Binding Site Prediction
Thi Lan Nguyen and colleagues report that gRASSP combines pretrained RNA representations with lightweight graph refinement for RNA-small molecule binding-site prediction.
bulk2scDiff: Generating Single-Cell Expression Profiles from Pseudobulk Data with Conditional Diffusion Models
Jiedan Xiao and colleagues report that a conditional diffusion model reconstructs plausible single-cell populations from bulk-like expression summaries.
A self-supervised DNA foundation model with collapse-resistant multimodal fusion
Yuyan Chen reports that a DNA foundation model combines sequence with local and global chromatin accessibility while trying to prevent one modality from overwhelming the representation.
Single-cell foundation models benefit from cross-modal training: adding proteomics data beats parameter scaling
Maximilien Burq and colleagues matched or beat much larger RNA-only models on most reported benchmarks using a small biological foundation model trained with proteomics.
AlphaConformers: Structure-guided sampling enables prediction of multiple protein conformations
Julie Liliane Daniel and colleagues report that alphaConformers uses structures from related proteins as template hypotheses to steer AlphaFold2 toward alternative conformations.
Mechanistic machine learning for prediction of prime editing outcomes
Alvin Hsu and colleagues report that optiPrime uses a mechanism-based machine-learning model to predict prime-editing outcomes and nominate edits intended to evade mismatch repair.
Virtual-cell verification enables self-auditing AI discovery for immune rejuvenation
Yue You and colleagues report that a preprint combines an immune-aging score, a tokenized virtual-cell model, and an agent that revises its own compound-scoring objective.
Products & tools
Related releases
Insilico Medicine
Insilico Virtual Aging Cell preview
Insilico Medicine launched an informational webpage and previewed a multi-agent virtual-cell concept that treats biological age as a conditioning variable across several biological scales.
- Access
- research preview
- License
- Not specified
Capital & partnerships
Related deals
platform agreement
Lundbeck
Lundbeck is expanding EVERSANA's AI-enabled platform across marketing, medical communications, market access, and media in its U.S. organization. The companies did not disclose financial terms or product-level performance.
- Deal value
- Undisclosed
- Lead investor
- Not stated
research collaboration
Chai Discovery
The companies announced a research collaboration using Chai's models for antibody discovery. Targets, disease areas, economics, timelines, and candidate-level results were not disclosed.
- Deal value
- Undisclosed
- Lead investor
- Not stated
research collaboration
Network Bio
Network Bio announced a collaboration combining its longitudinal cell-free RNA datasets with NVIDIA accelerated computing, Parabricks, and BioNeMo Recipes to build a foundation model called Nexus. Financial terms, development milestones, and external validation plans were not disclosed.
- Deal value
- Undisclosed
- Lead investor
- Not stated
series financing
Aureka Biotechnologies
Fund research and large-scale training of next-generation biological foundation models and upgrade the company's Lab-in-the-Loop platform.
- Deal value
- $100M
- Lead investor
- Granite Asia