Find open-source science resources

Python package for segmenting geospatial data with the Segment Anything Model (SAM), enabling zero-shot object segmentation in satellite and aerial imagery for remote sensing and Earth observation (MIT, 4k+ stars)

Scalable genomic analysis.

MCP server enabling spatial transcriptomics analysis via natural language, integrating 60+ methods including SpaGCN, Cell2location, LIANA+, CellRank for Visium, Xenium, MERFISH platforms

Microsoft AI for Good Lab's open-source biodiversity research hub providing AI models, edge devices, and tools for wildlife monitoring and conservation, including MegaDetector (camera trap animal detection), SPARROW (species recognition), PytorchWildlife (conservation AI toolkit), and bioacoustics analysis pipelines (1K+ stars)

Research coding benchmark curated by scientists with 338 subproblems across 16 subdomains (physics, math, materials, biology, chemistry), evaluating LLMs on realistic scientific programming tasks with gold-standard solutions (NeurIPS 2024)

This package provides an interface between HDF5 and R. HDF5's main features are the ability to store and access very large and/or complex datasets and a wide variety of metadata on mass storage (disk) through a completely portable file format. The rhdf5 package is thus suited for the exchange of large and/or complex datasets between R and other software package, and for letting R applications work on datasets that are larger than the available RAM.

Bring the power and flexibility of AnnData to the R ecosystem, allowing you to effortlessly manipulate and analyse your single-cell data. This package lets you work with backed h5ad and zarr files, directly access various slots (e.g. X, obs, var), or convert the data into SingleCellExperiment and Seurat objects.

Curated, accuracy-first collection of benchmarks for evaluating LLMs on scientific reasoning and discovery across mathematics, physics, chemistry, materials science, biology, and agentic science (subinium, 29+ stars, MIT License, 2026)

The HGVS Nomenclature is an internationally-recognized standard for the description of DNA, RNA and protein sequence variants. It is used to convey variants in clinical reports and to share variants in publications and databases. The HGVS Nomenclature is administered by the [HGVS Variant Nomenclature Committee (HVNC)](https://hgvs-nomenclature.org/stable/hvnc/) under the auspices of the [Human Genome Organization (HUGO)](https://hugo-int.org/).

GBScleanR is a package for quality check, filtering, and error correction of genotype data derived from next generation sequcener (NGS) based genotyping platforms. GBScleanR takes Variant Call Format (VCF) file as input. The main function of this package is `estGeno()` which estimates the true genotypes of samples from given read counts for genotype markers using a hidden Markov model with incorporating uneven observation ratio of allelic reads. This implementation gives robust genotype estimation even in noisy genotype data usually observed in Genotyping-By-Sequnencing (GBS) and similar methods, e.g. RADseq. The current implementation accepts genotype data of a diploid population at any generation of multi-parental cross, e.g. biparental F2 from inbred parents, biparental F2 from outbred parents, and 8-way recombinant inbred lines (8-way RILs) which can be refered to as MAGIC population.

Open source PEM (Proton Exchange Membrane) fuel cell simulation tool.

BIOSZEN is an open-source R package and modular Shiny application for reproducible analysis and visualization of experimental biological data from Excel or CSV files. It supports statistical testing, control-based normalization, replicate-aware quality control, and customizable scientific plots. Its microbial growth-curve module analyzes optical-density time series and automatically extracts quantitative growth parameters, including maximum specific growth rate (µMax), doubling time, lag time, maximum optical density (ODmax), time to maximum growth, area under the curve (AUC), and initial optical density (OD0). Results can be exported as processed datasets, statistical summaries, analysis metadata, scientific graphics, and editable PowerPoint figures.

Provide functions for retrieving, exploratory analyzing and visualizing the Human Protein Atlas data. HPAanalyze is designed to fullfill 3 main tasks: (1) Import, subsetting and export downloadable datasets; (2) Visualization of downloadable datasets for exploratory analysis; and (3) Working with the individual XML files. This package aims to serve researchers with little programming experience, but also allow power users to use the imported data as desired.

A DINOv2 ViT-S/14-reg fine-tuned so that an image of a molecular structure diagram embeds where its molecule embeds in the frozen MIST-28M embedding space. Objective: smooth-L1 regression onto the frozen target, no negatives (the JEPA move).

A DINOv2 ViT-S/14-reg fine-tuned so that an image of a molecular structure diagram embeds where its molecule embeds in the frozen MIST-28M embedding space. Objective: SigLIP sigmoid pairwise loss.

Text-space optimizer that treats agent skill documents as trainable parameters for frozen LLMs, using scored rollouts and held-out validation gates to iteratively improve reusable natural-language skills; includes SkillOpt-Sleep for nightly self-evolution and improves accuracy across Claude Code, Codex, Copilot, and direct-chat harnesses, making it a meta-tool for evolving scientific agent skill workflows (15.5K+ stars, MIT License, PyPI)

AI-powered note linking and research graph navigation

ReviewAid is an open-source AI-assisted tool for full-text screening and data extraction in systematic reviews. It supports evidence synthesis workflows by using large language models to classify articles according to user-defined PICO criteria and extract structured information from full-text publications. ReviewAid is designed as a supplementary reviewer rather than a replacement for human judgement. It aims to reduce manual workload, improve consistency, and assist researchers during screening and data extraction while maintaining human oversight throughout the evidence synthesis process.

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

NOS-TLPlot is an open-source tool for visualizing Newcastle–Ottawa Scale (NOS) risk-of-bias assessments in systematic reviews. It converts NOS star ratings into publication-ready traffic-light plots and 12 specialized visualizations, enabling reviewers and readers to interpret study-level risk-of-bias results clearly and reproducibly.

RejuvenationKit is an open-source Python toolkit for reproducible auditing and analysis of longitudinal preclinical rejuvenation studies. It provides protocol-aware missingness checks, experimental-confounding diagnostics, attrition and analysis-readiness profiling, covariance-aware multichannel change detection, sequential response monitoring, randomized longitudinal inference, visualization, and integrity-tracked report bundles.

Unified interface for local, global, gradient-based and derivative-free optimization (800+ stars)

This is a R package to compute the automorphisms between pairwise aligned DNA sequences represented as elements from a Genomic Abelian group. In a general scenario, from genomic regions till the whole genomes from a given population (from any species or close related species) can be algebraically represented as a direct sum of cyclic groups or more specifically Abelian p-groups. Basically, we propose the representation of multiple sequence alignments of length N bp as element of a finite Abelian group created by the direct sum of homocyclic Abelian group of prime-power order.

A vocabulary used in tandem with SHACL for representing node shapes

Evaluating the reliability of your own metrics and the measurements done on your own datasets by analysing the stability and goodness of the classifications of such metrics.

Aggregate results from bioinformatics analyses across many samples into a single report.

bettr provides a set of interactive visualization methods to explore the results of a benchmarking study, where typically more than a single performance measures are computed. The user can weight the performance measures according to their preferences. Performance measures can also be grouped and aggregated according to additional annotations.

Design, conduct and analyze results of AI-powered surveys and experiments. Simulate social science and market research with large numbers of AI agents and LLMs (460+ stars, 2024)

Python library for blazing-fast genomic interval operations and genomic file formats I/O on Polars DataFrames

The DCAT-AP conversion to a LinkML Schema is the intended point of truth for the DCAT-AP+ schema, but could be used alternatively as a LinkML representation of DCAT-AP for other Projects. It is a port of DCAT-AP to the LinkML world that is as faithful to the original as possible. This Persistent Identifier does not only provide the SHACL Shape, but could also be used as described [here](https://github.com/perma-id/w3id.org/tree/cecbc2e5f40d928f05ed5306d24fc60db0e7bb21/nfdi-de/dcat-ap-plus). DCAT-AP+ is a [LinkML](https://linkml.io/)-based extension of the [DCAT Application Profile 3.0](https://semiceu.github.io/DCAT-AP/releases/3.0.0/) that adds a provenance layer for describing how a dataset was generated and what it is about, using the [Starting Point Terms of PROV-O](https://www.w3.org/TR/prov-o/#description-starting-point-terms), the [QUDT ontology](https://www.qudt.org/), and [Dublin Core Terms](http://purl.org/dc/terms/).

Module for single-cell data extraction given a segmentation mask and multi-channel image.

A toolkit for visualizations in materials informatics.

Microsoft's foundation model for the Earth system supporting weather, air pollution, and ocean wave forecasting at multiple resolutions, trained on 1M+ hours of diverse atmospheric data (Nature 2025)

The main purpose of HTSlib is to provide access to genomic information files, both alignment data (SAM, BAM, and CRAM formats) and variant data (VCF and BCF formats). The library also provides interfaces to access and index genome reference data in FASTA format and tab-delimited files with genomic coordinates. It is utilized and incorporated into both SAMtools and BCFtools.

Provides a unified interface to a variety of GSEA techniques from different bioconductor packages. Results are harmonized into a single object and can be interrogated uniformly for quick exploration and interpretation of results. Interactive exploration of GSEA results is enabled through a shiny app provided by a sparrow.shiny sibling package.

Contrastively fine-tuned ESM-2 150M producing fixed-length protein embeddings where biological similarity maps to embedding proximity. Intended for retrieval, clustering, and nearest-neighbour transfer.

🩺 HealthGPT-Pro: A High-Performance Multimodal Large Language Model for Medical Understanding and Analysis

Ontology representation of the [International Committee on Taxonomy of Viruses (ICTV)](https://ictv.global/) for the [EVORA project](https://evora-project.eu/)

This ontology describes sensors, actuators and observations, and related concepts. It does not describe domain concepts, time, locations, etc. these are intended to be included from other ontologies via OWL imports.

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

Meta's comprehensive ML ecosystem for materials/chemistry with 118M+ DFT calculations, EquiformerV2 models achieving top Matbench Discovery performance

A binary healthy hard coral vs bleached hard coral classifier built on top of the ReefNet species LoRA model BobDerBaum/bioclip-2.5-vith14-reefnet-lora, which provides the fine-tuned vision-encoder LoRA adapters. Only a small linear head is trained on top (frozen backbone + LoRA + 2-way linear…

A LoRA fine-tune of imageomics/bioclip-2.5-vith14 trained contrastively on the ReefNet 1.0 coral-reef species dataset (ReefNet/ReefNet-1.0), 92-class global curated split (train 48,312 / image-val 32,792 / image-test 33,090 / source-val 8,074; split cache 56ea94e36f9f).

PyTorch-based embedding instance segmentation algorithm optimized for accurate, efficient, and portable cell and nucleus segmentation across fluorescence and brightfield microscopy images, achieving state-of-the-art speed and accuracy with lightweight model sizes suitable for edge deployment (224+ stars, Apache 2.0)