Find open-source science resources

Structure-aware protein language model using 3D structural vocabulary (Foldseek) for joint sequence-structure pretraining, achieving SOTA on protein engineering and fitness prediction benchmarks (ICML 2024, Westlake University & Repl)

Scientific foundation model and AI research copilot for idea generation, cross-disciplinary connection discovery, and hypothesis formation; trained with a decoupled reward-comment RL architecture and achieves GPT-4o-competitive novelty/rationale on STEM and social-science idea-generation benchmarks (270+ stars, MIT License, 2026)

Bilingual protein language model translating between protein sequence and structure, finetuned from ProtT5-XL on 17M AlphaFoldDB structures using Foldseek's 3Di structural alphabet, enabling sequence-to-structure prediction, structure-to-sequence inverse folding, and unified protein representation learning (RostLab, 310+ stars)

First benchmark for automatic video generation from scientific papers (NeurIPS 2025)

RFantibody is a pipeline for structure-based de novo antibody and nanobody design, integrating backbone design with RFdiffusion, sequence design with ProteinMPNN, and structure prediction with RoseTTAFold2. It provides a comprehensive toolset for generating and filtering high-quality antibody designs.

Interactive personal genome analysis toolkit using Claude Code and Python. Parses raw genotyping data from consumer DNA services and analyzes SNPs across 17 categories including health risks, pharmacogenomics, ancestry, and nutrition, with a terminal-style HTML dashboard.

Open-source platform for building, extending, and experimenting with scientific agents, providing modular agent construction tools and standardized evaluation pipelines for accelerating autonomous scientific discovery research (748+ stars, MIT License)

Our scLANE model uses truncated power basis spline models to build flexible, interpretable models of single cell gene expression over pseudotime or latent time. The modeling architectures currently supported are Negative-binomial GLMs, GEEs, & GLMMs. Downstream analysis functionalities include model comparison, dynamic gene clustering, smoothed counts generation, gene set enrichment testing, & visualization.

First benchmark evaluating LLMs' ability to rediscover scientific laws through interactive experimentation across 324 tasks in 12 physics domains, featuring memorization-resistant metaphysical shifts of canonical laws (HKUST)

Provides functionality for untargeted LC-MS metabolomics research as specified in the associated protocol article in the 'Metabolomics Data Processing and Data Analysis—Current Best Practices' special issue of the Metabolites journal (2020). This includes tabular data preprocessing and quality control, uni- and multivariate analysis as well as quality control visualizations, feature-wise visualizations and results visualizations. Raw data preprocessing and functionality related to biological context, such as pathway analysis, is not included.

Analyze and visualize Mutation Annotation Format (MAF) files from large scale sequencing studies. This package provides various functions to perform most commonly used analyses in cancer genomics and to create feature rich customizable visualzations with minimal effort.

Flow-matching protein folding model using only general-purpose transformer layers, scaled to 3B parameters and trained on 8.6M+ distilled structures; challenges the reliance on complex domain-specific architectures and supports PyTorch and MLX backends with model sizes from 100M to 3B parameters (985+ stars, MIT License)

FASTQ and SAM quality control using Python.

Provides means to interactively visualize guide RNAs (gRNAs) in GuideSet objects via Shiny application. This GUI can be self-contained or as a module within a larger Shiny app. The content of the app reflects the annotations present in the passed GuideSet object, and includes intuitive tools to examine, filter, and export gRNAs, thereby making gRNA design more user-friendly.

BWA-MEM drop-in replacement: 2-3x faster, 2-5x cheaper, 100% identical output on standard CPUs.

lumpy: a general probabilistic framework for structural variant discovery.

Provides visualization functionality for untargeted LC-MS metabolomics research. Includes quality control visualizations, feature-wise visualizations and results visualizations.

A script to run structural alerts using the RDKit and ChEMBL

sRACIPE implements a randomization-based method for gene circuit modeling. It allows us to study the effect of both the gene expression noise and the parametric variation on any gene regulatory circuit (GRC) using only its topology, and simulates an ensemble of models with random kinetic parameters at multiple noise levels. Statistical analysis of the generated gene expressions reveals the basin of attraction and stability of various phenotypic states and their changes associated with intrinsic and extrinsic noises. sRACIPE provides a holistic picture to evaluate the effects of both the stochastic nature of cellular processes and the parametric variation.

A polymorphic bayesian genotyping model with wide applicability.

Physics-informed neural networks

Azure Semantic Kernel multi-agent PPT generation reference

Uniparental disomy (UPD) is a genetic condition where an individual inherits both copies of a chromosome or part of it from one parent, rather than one copy from each parent. This package contains a HMM for detecting UPDs through HTS (High Throughput Sequencing) data from trio assays. By analyzing the genotypes in the trio, the model infers a hidden state (normal, father isodisomy, mother isodisomy, father heterodisomy and mother heterodisomy).

Multi-agent system automatically transforming research papers into interactive AI agents with MCP server generation, tutorial auto-detection, and benchmark extraction (2.2K+ stars, MIT License, 2025)

Sequencing and microarray samples often are collected or processed in multiple batches or at different times. This often produces technical biases that can lead to incorrect results in the downstream analysis. BatchQC is a software tool that streamlines batch preprocessing and evaluation by providing interactive diagnostics, visualizations, and statistical analyses to explore the extent to which batch variation impacts the data. BatchQC diagnostics help determine whether batch adjustment needs to be done, and how correction should be applied before proceeding with a downstream analysis. Moreover, BatchQC interactively applies multiple common batch effect approaches to the data and the user can quickly see the benefits of each method. BatchQC is developed as a Shiny App. The output is organized into multiple tabs and each tab features an important part of the batch effect analysis and visualization of the data. The BatchQC interface has the following analysis groups: Summary, Differential Expression, Median Correlations, Heatmaps, Circular Dendrogram, PCA Analysis, Shape, ComBat and SVA.

scQTLtools is a comprehensive R/Bioconductor package that facilitates end-to-end single-cell eQTL analysis, from preprocessing to visualization

This package draws protein schematics from Uniprot API output. From the JSON returned by the GET command, it creates a dataframe from the Uniprot Features API. This dataframe can then be used by geoms based on ggplot2 and base R to draw protein schematics.

Evaluating multimodal autonomous agents in realistic scientific workflows across real scientific software environments (KAlgebra, Celestia, Grass GIS, Lean 4, etc.) with VM-based evaluation infrastructure and agent trajectories

Subgroup classification is a basic task in genomic data analysis, especially for gene expression and DNA methylation data analysis. It can also be used to test the agreement to known clinical annotations, or to test whether there exist significant batch effects. The cola package provides a general framework for subgroup classification by consensus partitioning. It has the following features: 1. It modularizes the consensus partitioning processes that various methods can be easily integrated. 2. It provides rich visualizations for interpreting the results. 3. It allows running multiple methods at the same time and provides functionalities to straightforward compare results. 4. It provides a new method to extract features which are more efficient to separate subgroups. 5. It automatically generates detailed reports for the complete analysis. 6. It allows applying consensus partitioning in a hierarchical manner.

The core functionality of the package is to provide coordinates of genes on the BioCarta pathway images and to provide methods to add self-defined graphics to the genes of interest.

A new clustering algorithm, "binary cut", for clustering similarity matrices of functional terms is implemeted in this package. It also provides functions for visualizing, summarizing and comparing the clusterings.

This package can easily make heatmaps which are produced by the ComplexHeatmap package into interactive applications. It provides two types of interactivities: 1. on the interactive graphics device, and 2. on a Shiny app. It also provides functions for integrating the interactive heatmap widgets for more complex Shiny app development.

GREAT (Genomic Regions Enrichment of Annotations Tool) is a type of functional enrichment analysis directly performed on genomic regions. This package implements the GREAT algorithm (the local GREAT analysis), also it supports directly interacting with the GREAT web service (the online GREAT analysis). Both analysis can be viewed by a Shiny application. rGREAT by default supports more than 600 organisms and a large number of gene set collections, as well as self-provided gene sets and organisms from users. Additionally, it implements a general method for dealing with background regions.

This package implements infrastructures for ontology analysis by offering efficient data structures, fast ontology traversal methods, and elegant visualizations. It provides a robust toolbox supporting over 70 methods for semantic similarity analysis.

Genome level Trellis graph visualizes genomic data conditioned by genomic categories (e.g. chromosomes). For each genomic category, multiple dimensional data which are represented as tracks describe different features from different aspects. This package provides high flexibility to arrange genomic categories and to add self-defined graphics in the plot.

Enriched heatmap is a special type of heatmap which visualizes the enrichment of genomic signals on specific target regions. Here we implement enriched heatmap by ComplexHeatmap package. Since this type of heatmap is just a normal heatmap but with some special settings, with the functionality of ComplexHeatmap, it would be much easier to customize the heatmap as well as concatenating to a list of heatmaps to show correspondance between different data sources.

Open-source toolkit and benchmark for learning-based theorem proving in Lean, providing programmatic Lean interaction, a 98K+ theorem dataset extracted from 217 Lean projects, and ReProver—the first retrieval-augmented LLM-based theorem prover for Lean—with reproducible training pipelines underpinning much subsequent Lean prover research (Caltech & NVIDIA, NeurIPS 2023 Outstanding Paper, Datasets & Benchmarks)

The package comprises a set of pretrained machine learning models to predict basic immune cell types. This enables all users to quickly get a first annotation of the cell types present in their dataset without requiring prior knowledge. scAnnotatR also allows users to train their own models to predict new cell types based on specific research needs.

Design primers for targeted single-cell RNA-seq used by TAP-seq. Create sequence templates for target gene panels and design gene-specific primers using Primer3. Potential off-targets can be estimated with BLAST. Requires working installations of Primer3 and BLASTn.

Discrete diffusion framework for generative protein sequence design over evolutionary-scale databases, supporting unconditional generation, evolutionary-guided conditional design, motif scaffolding, and intrinsically disordered region generation through order-agnostic autoregressive diffusion, enabling sequence-only protein design without structural priors (Microsoft Research, Nature Communications 2024)

CebraEM is a bioinformatics tool for analyzing and processing large-scale imaging data, providing a pipeline for segmentation, annotation, and analysis with support for both Linux and Windows environments. It includes modules for core functionality, annotation, and network analysis, requiring specific dependencies and a conda environment for execution.

A batteries-included toolkit for the GPU-accelerated OpenMM molecular simulation engine.

A seamless interface to the MEME Suite family of tools for motif analysis. 'memes' provides data aware utilities for using GRanges objects as entrypoints to motif analysis, data structures for examining & editing motif lists, and novel data visualizations. 'memes' functions and data structures are amenable to both base R and tidyverse workflows.

`tidyCoverage` framework enables tidy manipulation of collections of genomic tracks and features using `tidySummarizedExperiment` methods. It facilitates the extraction, aggregation and visualization of genomic coverage over individual or thousands of genomic loci, relying on `CoverageExperiment` and `AggregatedCoverage` classes. This accelerates the integration of genomic track data in genomic analysis workflows.

A library for building, manipulating, analyzing and automatic design of molecules, including a genetic algorithm.

AI-human collaborative research platform where a human researcher works with a team of LLM agents via team and individual meetings to perform scientific research; demonstrated by designing new SARS-CoV-2 nanobodies with wet-lab validation

Deep equivariant generative model predicting ligand-specific protein-ligand complex structures with dynamic receptor conformational flexibility, enabling accurate docking for flexible protein targets

Cross-modal self-supervised foundation model for galaxies by Polymathic AI, jointly embedding multi-band galaxy imaging and optical spectra into a shared latent space to enable zero/few-shot redshift estimation, galaxy property prediction, morphology classification, and cross-modal similarity search (MNRAS Letters 2024)

Generative AI framework for inverse design of 3D RNA structure and function using geometric deep learning, learning design rules from 3D structures to capture complex tertiary interactions (pseudoknots, non-canonical base pairs) with expert-level accuracy for designing functional RNAs including aptamers and ribozymes (bioRxiv 2025)

Gene signatures of TB progression, TB disease, and other TB disease states have been validated and published previously. This package aggregates known signatures and provides computational tools to enlist their usage on other datasets. The TBSignatureProfiler makes it easy to profile RNA-Seq data using these signatures and includes common signature profiling tools including ASSIGN, GSVA, and ssGSEA. Original models for some gene signatures are also available. A shiny app provides some functionality alongside for detailed command line accessibility.