
by BioMap
scGPT is a foundation model for single-cell multi-omics built with generative AI. The repository has the code, pretrained checkpoints and tutorials.
Web appFree
Step 1: Install scGPT with pip, optionally with flash-attention.
Step 2: Download a pretrained checkpoint folder, such as the whole-human one.
Step 3: Load it with the load_pretrained function.
Step 4: Fine-tune it on your data with the integration example, or run the zero-shot tutorials.
Step 5: Or open the online apps in your browser for mapping, annotation or GRN inference.
The README recommends the whole-human model for most applications, and organ-specific models can compete when your data matches their tissue.
No. It is optional, and the weights load on CPU, GPU or flash-attn backends.
Yes. Online apps for reference mapping, cell annotation and GRN inference run in the browser on cloud GPUs.
The README says searching for 10,000 query cells takes under one second on a GPU, with an index of over 33 million cells under 1GB.
ProtGPT2Generative language model for producing protein sequences with natural protein-like characteristics
MatterSimMatterSim is a deep learning atomistic model that works across elements, temperatures and pressures. It is distributed as a Python package with pre-trained models.