Find open-source science resources

Graph neural network operating entirely at the atomic level for protein-ligand conformational ensemble prediction and docking, generating diverse solutions through rapid stochastic denoising to model conformational heterogeneity (Baker Lab, bioRxiv 2025)

Teaching Large Language Models the Language of Biology through single-cell transcriptomics (ICML 2024)

MedQA Github porject: https://github.com/hussien/MedQA Read the detailed expermintal report here: https://github.com/hussien/MedQA/blob/main/report/MedQA.pdf

This package primarily identifies variants in mitochondrial genomes from BAM alignment files. It filters these variants to remove RNA editing events then estimates their evolutionary relationship (i.e. their phylogenetic tree) and groups single cells into clones. It also visualizes the mutations and providing additional genomic context.

Use this package to interface with the WikiPathways API. It provides programmatic access to WikiPathways content in multiple data and image formats, including official monthly release files and convenient GMT read/write functions.

ChemFIE-BED is a sentence-transformers based on gbyuvd/chemselfies-base-bertmlm fine-tuned on around (for now) 2 million pairs of valid molecules' SELFIES (Krenn et al. 2020) taken from COCONUTDB (Sorokina et al. 2021) and ChemBL34 (Zdrazil et al. 2023).

Family of codon-resolution language models trained on 130 million protein-coding sequences from over 20,000 species, enabling cross-species gene expression prediction and codon-level functional genomics (2025)

ChemFormula provides a class for working with chemical formulas. It allows parsing chemical formulas, calculating formula weights, and generating formatted output strings (e.g. in HTML, LaTeX, or Unicode).

End-to-end deep learning approach for RNA tertiary structure prediction with a flexible nucleobase center representation, achieving ~7 Å C1' RMSD across test RNAs and predicting ~545,000 structures covering 2,200+ RNA families (Kihara Lab, Purdue University, 50+ stars)

GitHub homepage: Cell2Sentence GitHub

This package provides a novel method for interpreting new transcriptomic datasets through near-instantaneous comparison to public archives without high-performance computing requirements. Through the pre-computed index, users can identify public resources associated with their dataset such as gene sets, MeSH term, and publication. Functions to identify interpretable annotations and intuitive visualization options are implemented in this package.

Discovering interpretable features in protein language models via sparse autoencoders, enabling mechanistic understanding of PLM representations for protein engineering and design (288+ stars, MIT License)

First versatile medical reasoning agent for chest X-ray interpretation, dynamically integrating state-of-the-art CXR analysis tools and multimodal LLMs into a unified framework; introduces ChestAgentBench with 2,500 complex medical queries across 7 categories (bowang-lab, 1.1K+ stars)

A library for computational chemistry (DFT) for input file generation, data extraction, method screening and analysis.

A controlled vocabulary to support the study of transcription in the mouse brain

S3segmenter is a Matlab-based set of functions that generates single cell (nuclei and cytoplasm) label masks.

Experiments with expanded ensembles to explore chemical space.

TENET identifies key transcription factors (TFs) and regulatory elements (REs) linked to a specific cell type by finding significantly correlated differences in gene expression and RE DNA methylation between case and control input datasets, and identifying the top genes by number of significant RE DNA methylation site links. It also includes many tools for visualization and analysis of the results, including plots displaying and comparing methylation and expression data and methylation site link counts, survival analysis, TF motif searching in the vicinity of linked RE DNA methylation sites, custom TAD and peak overlap analysis, and UCSC Genome Browser track file generation. A utility function is also provided to download methylation, expression, and patient survival data from The Cancer Genome Atlas (TCGA) for use in TENET or other analyses.

Conversational data analysis using natural language

The tidyomics ecosystem is a set of packages for ’omic data analysis that work together in harmony; they share common data representations and API design, consistent with the tidyverse ecosystem. The tidyomics package is designed to make it easy to install and load core packages from the tidyomics ecosystem with a single command.

This model is a fine-tuned version of google/medgemma-4b-it adapted for binary mammogram classification on the OMAMA 256×256 dataset. The dataset consists of ~154k mammogram image slices (.npz) with metadata JSONs providing labels (NonCancer, Cancer).

Statistical methods for detection of differential splicing (differential exon usage) in RNA-seq and exon microarray data, using L1-regularization (lasso) to improve power.

Provides S4 classes for general nucleases, CRISPR nucleases, CRISPR nickases, and base editors.Several CRISPR-specific genome arithmetic functions are implemented to help extract genomic coordinates of spacer and protospacer sequences. Commonly-used CRISPR nuclease objects are provided that can be readily used in other packages. Both DNA- and RNA-targeting nucleases are supported.

MetaDICT is a method for the integration of microbiome data. This method is designed to remove batch effects and preserve biological variation while integrating heterogeneous datasets. MetaDICT can better avoid overcorrection when unobserved confounding variables are present.

Combine generalised least squares methodology from the nlme package for dealing with autocorrelation with penalised least squares methods from the glmnet package to deal with high dimensionality. This pengls packages glues them together through an iterative loop. The resulting method is applicable to high dimensional datasets that exhibit autocorrelation, such as spatial or temporal data.

For a convenient overview and download list, visit our model page for this model.

F-informed MDS is a new multidimensional scaling-based ordination method that configures data distribution based on the F-statistic (i.e., the ratio of dispersion between groups with shared or differing labels).

A quasi-simulation based approach to performing power analysis for EWAS (Epigenome-wide association studies) with continuous or binary outcomes. 'EpipwR' relies on empirical EWAS datasets to determine power at specific sample sizes while keeping computational cost low. EpipwR can be run with a variety of standard statistical tests, controlling for either a false discovery rate or a family-wise type I error rate.

Tahoe-x1 is a family of perturbation-trained single-cell foundation models with up to 3 billion parameters, developed by Tahoe Therapeutics. Pretrained on 266 million single-cell transcriptomic profiles including the Tahoe-100M perturbation compendium, Tahoe-x1 achieves state-of-the-art performance…

BioCAP is a foundation model for biology organismal images. It is trained on TreeOfLife-10M with synthetic captions (TreeOfLife-10M-Captions) as supervision on the basis of a CLIP model (ViT-B/16) pre-trained by OpenAI. BioCAP achieves state-of-the-art performance on text-image retrieval tasks.

AI-powered pipeline converting papers into interactive websites, posters, and multimedia presentations with "Let's Make Your Paper Alive!" philosophy

Geometry Aware Operator Transformer serving as an efficient and accurate neural surrogate for PDEs on arbitrary domains, combining geometric priors with transformer architectures for scientific computing (ETH Zurich CAMLab, 92+ stars)

anglemania extracts genes from multi-batch scRNA-seq experiments for downstream dataset integration. It shows improvement over the conventional usage of highly-variable genes for many integration tasks. We leverage gene-gene correlations that are stable across batches to identify biologically informative genes which are less affected by batch effects. Currently, its main use is for single-cell RNA-seq dataset integration, but it can be applied for other multi-batch downstream analyses such as NMF.

The Nucleotide Transformers are a collection of foundational language models that were pre-trained on DNA sequences from whole-genomes. Compared to other approaches, our models do not only integrate information from single reference genomes, but leverage DNA sequences from over 3,200 diverse human…

mitology allows to study the mitochondrial activity throught high-throughput RNA-seq data. It is based on a collection of genes whose proteins localize in to the mitochondria. From these, mitology provides a reorganization of the pathways related to mitochondria activity from Reactome and Gene Ontology. Further a ready-to-use implementation of MitoCarta3.0 pathways is included.

Spatially-aware quality control (QC) software for both spot-level and artifact-level QC in spot-based spatial transcripomics, such as 10x Visium. These methods calculate local (nearest-neighbors) mean and variance of standard QC metrics (library size, unique genes, and mitochondrial percentage) to identify outliers spot and large technical artifacts.

omicsGMF is a Bioconductor package that uses the sgdGMF-framework of the \code{sgdGMF} package for highly performant and fast matrix factorization that can be used for dimensionality reduction, visualization and imputation of omics data. It considers data from the general exponential family as input, and therefore suits the use of both RNA-seq (Poisson or Negative Binomial data) and proteomics data (Gaussian data). It does not require prior transformation of counts to the log-scale, because it rather optimizes the deviances from the data family specified. Also, it allows to correct for known sample-level and feature-level covariates, therefore enabling visualization and dimensionality reduction upon batch correction. Last but not least, it deals with missing values, and allows to impute these after matrix factorization, useful for proteomics data. This Bioconductor package allows input of SummarizedExperiment, SingleCellExperiment, and QFeature classes.

UMI-4C is a technique that allows characterization of 3D chromatin interactions with a bait of interest, taking advantage of a sonication step to produce unique molecular identifiers (UMIs) that help remove duplication bias, thus allowing a better differential comparsion of chromatin interactions between conditions. This package allows processing of UMI-4C data, starting from FastQ files provided by the sequencing facility. It provides two statistical methods for detecting differential contacts and includes a visualization function to plot integrated information from a UMI-4C assay.

Our R package MultiRNAflow provides an easy to use unified framework allowing to automatically make both unsupervised and supervised (DE) analysis for datasets with an arbitrary number of biological conditions and time points. In particular, our code makes a deep downstream analysis of DE information, e.g. identifying temporal patterns across biological conditions and DE genes which are specific to a biological condition for each time.

Adopting tipping-point theory to transcriptome profiles to unravel disease regulatory trajectory.

`SPOTlight` provides a method to deconvolute spatial transcriptomics spots using a seeded NMF approach along with visualization tools to assess the results. Spatially resolved gene expression profiles are key to understand tissue organization and function. However, novel spatial transcriptomics (ST) profiling techniques lack single-cell resolution and require a combination with single-cell RNA sequencing (scRNA-seq) information to deconvolute the spatially indexed datasets. Leveraging the strengths of both data types, we developed SPOTlight, a computational tool that enables the integration of ST with scRNA-seq data to infer the location of cell types and states within a complex tissue. SPOTlight is centered around a seeded non-negative matrix factorization (NMF) regression, initialized using cell-type marker genes and non-negative least squares (NNLS) to subsequently deconvolute ST capture locations (spots).

CaMutQC is able to filter false positive mutations generated due to technical issues, as well as to select candidate cancer mutations through a series of well-structured functions by labeling mutations with various flags. And a detailed and vivid filter report will be offered after completing a whole filtration or selection section. Also, CaMutQC integrates serveral methods and gene panels for Tumor Mutational Burden (TMB) estimation.

This package is designed for calling lineage-informative mitochondrial mutations using single-cell sequencing data, such as scRNASeq and scATACSeq (preferably the latter due to RNA editing issues). It includes functions for mutation calling and visualization. Mutation calling is done using beta-binomial distribution.

Specialized model for Species Entity Recognition - Species and organism names

GeneJEPA is a Joint-Embedding Predictive Architecture (JEPA) trained for self-supervised representation learning on scRNA-seq. It uses a Perceiver-style encoder to handle sparse, high-dimensional gene count vectors and a Fourier-feature tokenizer for numerical tokenization.

A Package For Training SNAP Interatomic Potentials for use in the LAMMPS molecular dynamics package.

Methods and models for handling zero-inflated single cell assay data.

Uses segmented copy number data to estimate tumor cell percentage and produce copy number plots displaying absolute copy numbers.