Find open-source science resources

Machine learning software for extracting structured metadata from scholarly documents

Official MathWorks toolkit connecting AI agents to MATLAB via the MATLAB MCP Server and curated skills, enabling trusted engineering and scientific computing workflows with idiomatic code generation, testing, and error diagnosis in Claude Code, GitHub Copilot, OpenAI Codex, and Gemini CLI (686+ stars, BSD-3-Clause, 2026)

Python toolkit for fine-tuning geospatial foundation models

Motivation: The understanding of cancer mechanism requires the identification of genes playing a role in the development of the pathology and the characterization of their role (notably oncogenes and tumor suppressors). Results: We present an R/bioconductor package called MoonlightR which returns a list of candidate driver genes for specific cancer types on the basis of TCGA expression data. The method first infers gene regulatory networks and then carries out a functional enrichment analysis (FEA) (implementing an upstream regulator analysis, URA) to score the importance of well-known biological processes with respect to the studied cancer type. Eventually, by means of random forests, MoonlightR predicts two specific roles for the candidate driver genes: i) tumor suppressor genes (TSGs) and ii) oncogenes (OCGs). As a consequence, this methodology does not only identify genes playing a dual role (e.g. TSG in one cancer type and OCG in another) but also helps in elucidating the biological processes underlying their specific roles. In particular, MoonlightR can be used to discover OCGs and TSGs in the same cancer type. This may help in answering the question whether some genes change role between early stages (I, II) and late stages (III, IV) in breast cancer. In the future, this analysis could be useful to determine the causes of different resistances to chemotherapeutic treatments.

Graph deep learning library for materials science powering the M3GNet universal interatomic potential across periodic-table elements, with property prediction, structure relaxation, and crystal generation workflows built on PyTorch and DGL (576+ stars, BSD-3-Clause, actively maintained)

Production-grade ETL for transforming complex documents into structured formats, with open-source API

Research agent system deeply integrated with Zotero supporting Agent Mode, skills, multi-model backends (OpenAI-compatible, Claude Code, WebChat, Codex), and MinerU PDF parsing for literature Q&A, summarization, figure inspection, and source comparison (1.3K+ stars, 2026)

Data management of large-scale whole-genome sequencing variant calls with thousands of individuals: genotypic data (e.g., SNVs, indels and structural variation calls) and annotations in SeqArray GDS files are stored in an array-oriented and compressed manner, with efficient data access using the R programming language.

Pretrained time series foundation model for long-horizon forecasting across diverse scientific domains including climate variables, biomedical signals, and physical observations; decoder-only Transformer architecture with strong zero-shot generalization (19.8K+ stars, Apache 2.0, 2024-2025)

Local-first, open-source healthcare AI toolkit for clinical NLP and PHI/PII de-identification across 12 languages, running entirely on-device with 1,000+ specialized medical models; provides Python SDK, REST API, Docker deployment, and native Swift apps via OpenMedKit with Apple MLX/CoreML acceleration, supporting HIPAA-aware de-identification with 247 PII checkpoints (3K+ stars, Apache 2.0, arXiv 2508.01630)

Probabilistic programming

A large RDF store built from American governmental data. This semantic space has a mixture of direct terms and subspaces.

Graph neural network library for PyTorch enabling molecular modeling, materials discovery, protein interaction networks, and scientific knowledge graph learning (23.7k+ stars)

Open-source 3D architectural editor with an AI design assistant; build floor plans with walls, doors, windows, and furniture using natural language, with real-time WebGPU-powered previews (TangSY, 59+ stars, MIT License, 2026)

The ESRFET ontology describes beamline experimental techniques at ESRF to enhance the FAIRness of experimental metadata. It supports semantic interoperability and reasoning over relationships between techniques. (from homepage)

First any-to-any generative foundation model for Earth Observation, enabling unified multimodal understanding and generation across diverse satellite sensors and geospatial tasks through a single architecture (258+ stars)

nnU-Net is a self-configuring method for deep learning-based biomedical image segmentation, developed by the Applied Computer Vision Lab (ACVL) of Helmholtz Imaging and the Division of Medical Image Computing at the German Cancer Research Center (DKFZ). It is designed to automatically adapt to a given dataset, analyzing the provided training cases to configure a matching U-Net-based segmentation pipeline without requiring expertise from the user. The tool provides pretrained models for Pancreas and Pancreas tumor segmentation, Colon cancer primaries segmentation, Abdominal organ segmentation, Liver and liver tumor segmentation, Kidney and kidney tumor segmentation, Brain Tumor segmentation and Hippocampus (MR data) segmentation

This tutorial aims to illustrate the process of setting up a simulation system containing a protein in complex with a ligand, step by step, using the BioExcel Building Blocks library (biobb). The particular example used is the T4 lysozyme L99A/M102Q protein (PDB code 3HTB), in complex with the 2-propylphenol small molecule (3-letter Code JZ4).

An extension of Schema.org to annotate metadata on software projects

An ultrafast protein aligner for `blastp` and `blastx` like searches.

scider is an user-friendly R package providing functions to model the global density of cells in a slide of spatial transcriptomics data. All functions in the package are built based on the SpatialExperiment object, allowing integration into various spatial transcriptomics-related packages from Bioconductor. After modelling density, the package allows for several downstream analysis, including colocalization analysis, boundary detection analysis and differential density analysis.

Lightweight Markdown-only skills for autonomous ML research with cross-model review loops, idea discovery, and experiment automation; no framework lock-in, works with Claude Code, Codex, OpenClaw, or any LLM agent (12.8K+ stars, MIT License, 2026)

Biomedical Model Context Protocol (MCP) server unifying literature search across PubMed/Europe PMC, entity pivoting across genes/variants/drugs/diseases/pathways/proteins, local study analytics, and Claude Code/Codex integration for agentic biomedical research (531+ stars, MIT License, 2025-2026)

AI coding assistant for JupyterLab with agent mode, supporting arbitrary LLM providers (2025+)

The information resource registry is a listing of data sources present in the NCATS Data Translator system. Each information resource has an identifier, a short description, and a URL to more information about that resource.

Beyond text-to-slides generation with PPTEval multi-dimensional evaluation (EMNLP 2025)

Chemical 2D structure editor application/applet based on the [Chemistry Development Kit](https://sourceforge.net/projects/cdk/).

PanAbyss is a tool for exploring and visualizing pangenome graphs. It allows users to search for and display regions of a pangenome using coordinates on a reference individual or based on annotations. It also enables searching for regions associated with a selected set of individuals (for example, those linked to a phenotype), computing proximity trees, and retrieving sequences from a given region.

This tutorial involves the use of a multilayer AutoEncoder (AE) for feature extraction and pattern recognition by analyzing Molecular Dynamic Simulations, step by step, using the BioExcel Building Blocks library (biobb)

Multi-LLM consensus framework for automated cell type annotation in single-cell transcriptomics, integrating predictions from 10+ large language models with iterative discussion and uncertainty quantification to reduce single-model biases, achieving up to 95% accuracy without reference datasets; available as CRAN R package and PyPI Python package with Scanpy/Seurat integration (2025)

The scRNAseqApp is a Shiny app package designed for interactive visualization of single-cell data. It is an enhanced version derived from the ShinyCell, repackaged to accommodate multiple datasets. The app enables users to visualize data containing various types of information simultaneously, facilitating comprehensive analysis. Additionally, it includes a user management system to regulate database accessibility for different users.

Python Library for Automating Molecular Simulation: input preparation, job execution, file management, output processing and building data workflows.

Machine learning toolkit for many-body quantum systems, implementing neural quantum states, variational Monte Carlo, and tensor network algorithms to solve ground-state and dynamical problems in condensed matter physics and quantum chemistry (EPFL & collaborators, Nature Physics 2019/2022+, 670+ stars)

MetaboDynamics is an R-package that provides a framework of probabilistic models to analyze longitudinal metabolomics data. It enables robust estimation of mean concentrations despite varying spread between timepoints and reports differences between timepoints as well as metabolite specific dynamics profiles that can be used for identifying "dynamics clusters" of metabolites of similar dynamics. Provides probabilistic over-representation analysis of KEGG functional modules and pathways as well as comparison between clusters of different experimental conditions.

A Molecular Interaction-Guided Graph Learning Framework for Multi-Omics Cancer Classification

An automated pipeline for the detection, integration and reporting of predefined features across a large number of mass spectrometry data files. It enables the real time annotation of multiple compounds in a single file, or the parallel annotation of multiple compounds in multiple files. A graphical user interface as well as command line functions will assist in assessing the quality of annotation and update fitting parameters until a satisfactory result is obtained.

Drop-in replacement for BiocNeighbors::findKNN using the jvecfor Java library, which builds on the jvector library to leverage the Java Vector API for portable SIMD acceleration across AVX2, AVX-512, and ARM NEON hardware. jvecfor/jvector implements HNSW-DiskANN approximate search and VP-tree exact search. The package achieves approximately 2x speedup over Annoy-based search at n >= 50K cells while returning output structurally identical to BiocNeighbors, making it suitable for seamless integration into existing Bioconductor single-cell workflows. Convenience wrappers delegate shared nearest-neighbor (SNN) and k-nearest-neighbor (KNN) graph construction to the bluster package.

Shared multimodal AI agent layer for geospatial Python packages (leafmap, geoai, geemap, STAC, NASA Earthdata) and QGIS, exposing geospatial tools to LLMs with structured metadata, confirmation hooks, and support for OpenAI, Anthropic, Google Gemini, Ollama, and more; includes the OpenGeoAgent QGIS plugin (456+ stars, MIT License)

Visualisation of peptide isotopic peaks and SIP peptide spectra match (PSM). Filtration of high quality PSM. Accurate isotopic abundance calculation of peptide and metabolites. Visualisation of SIP proteomics results.

Unified pre-trained model for general physics simulation via lifted geometric pre-training, augmenting static geometry with synthetic dynamics to enable dynamics-aware self-supervision without physics labels; improves industrial-fidelity benchmarks spanning fluid mechanics and solid mechanics while reducing labeled data requirements by 20–60% (Physics-Scaling, 224+ stars)

DNA methylation is generally considered to be associated with transcriptional silencing. However, comprehensive, genome-wide investigation of this relationship requires the evaluation of potentially millions of correlation values between the methylation of individual genomic loci and expression of associated transcripts in a relatively large numbers of samples. Methodical makes this process quick and easy while keeping a low memory footprint. It also provides a novel method for identifying regions where a number of methylation sites are consistently strongly associated with transcriptional expression. In addition, Methodical enables housing DNA methylation data from diverse sources (e.g. WGBS, RRBS and methylation arrays) with a common framework, lifting over DNA methylation data between different genome builds and creating base-resolution plots of the association between DNA methylation and transcriptional activity at transcriptional start sites.

Foundation model for tabular data that predicts on unseen real-world tables in a single forward pass, achieving accurate small-data classification and regression without task-specific training; widely applicable to scientific datasets with limited samples (7.4K+ stars, 2022-2026)

EMMO is a multidisciplinary effort to develop a standard representational framework (the ontology) for applied sciences. It is based on physics, analytical philosophy and information and communication technologies. It has been instigated by materials science to provide a framework for knowledge capture that is consistent with scientific principles and methodologies. (from GitHub)

A representation of variables appearing in models in the environmental research space.

Cheminformatics toolkit

This ontology models classes and relationships describing deep learning networks, their component layers and activation functions, as well as potential biases.

The primary goal of this ontology is to standardize the representation of molecular simulation data, processes, and methodologies across disparate simulation platforms, engines (e.g., GROMACS, AMBER, NAMD), and analysis tools, while ensuring these terms are interoperable with existing life sciences ontologies

PyTorch toolkit for deep neural networks in atomistic simulations, implementing SchNet, DimeNet++, PaiNN, and GemNet for molecular dynamics and quantum chemistry (900+ stars)

Analysis of molecular dynamics trajectories.

Agent-agnostic research infrastructure providing AI agents with a structured scientific workspace for deep PDF parsing, hybrid semantic/keyword literature search, citation-graph analysis, topic discovery, and academic writing workflows; natively integrates with Claude Code, Codex, Cursor, Cline, and AgentSkills.io (530+ stars, MIT License, 2026)