AlphaFold is an open‑source deep learning system for high‑accuracy protein structure prediction (monomers and multimers), producing ranked PDB models with per‑residue confidence (pLDDT), PAE/pTM estimates and machine‑readable outputs for downstream analysis.
Basenji is an open-source toolkit that trains deep convolutional neural networks on chromosome-scale DNA to predict quantitative regulatory activity, score variants (SAD/SED), and map distal regulatory elements and nucleotide drivers.
BPNet is a Python package and CLI for training base-resolution deep neural networks on functional genomics data (e.g., ChIP-nexus/ChIP-seq) and interpreting sequence motifs, their genomic locations, and motif interactions.
Cellpose is an open‑source, generalist deep‑learning tool for segmentation of cells, nuclei and other biological objects in 2D and 3D images, offering GUI/CLI/API access, pretrained models and options for fine‑tuning and large‑scale processing.
ColabFold brings state-of-the-art protein structure prediction to everyone: easy Google Colab notebooks and local tools that combine AlphaFold/RoseTTAFold with fast MMseqs2 MSA searches, batch/GPU workflows, and exports for downstream tools.
DeepRank converts protein–protein interfaces into 3D volumetric grids of structural and physicochemical features and provides pipelines to train CNNs for scoring, classifying, or regressing docking models; stores data in HDF5 and integrates with PyTorch.
DeepVariant is a deep learning–based variant calling pipeline that converts aligned reads (BAM/CRAM) into pileup tensors, uses a convolutional neural network to call SNPs and small indels, and produces high‑accuracy VCF/gVCF outputs across sequencing platforms.
Enformer is a transformer-based deep learning model that predicts genome-wide gene expression and epigenomic tracks from long-range DNA sequence, improving variant effect and enhancer–promoter inference for fine-mapping and regulatory interpretation.
ESM Atlas is an open, large-scale catalog of metagenomic protein structure predictions and pretrained protein language models (ESM family) that enable fast structure prediction, embeddings, sequence design, and large-scale search via web, API, and downloadable resources.
ESM-2 is Meta FAIR’s family of transformer protein language models that produce high-quality sequence embeddings and enable state-of-the-art single-sequence structure prediction (via ESMFold), contact maps, and downstream design and variant analyses.
ESM-IF1 is Meta FAIR's inverse-folding model that predicts or designs protein sequences from backbone coordinates, enabling fixed-backbone sequence design, conditional scoring of variants, and sequence sampling for partial or full structures.
ESMFold is a fast, end-to-end protein structure predictor that uses the ESM‑2 protein language model to generate atomic‑level 3D structures directly from sequence, enabling high‑throughput folding and exploration of metagenomic proteins.
