Find open-source science resources

![Language: English]()

A Chinese medical reasoning model fine-tuned from Qwen3.5-4B using a two-stage training pipeline: Supervised Fine-Tuning (SFT) for format alignment, followed by Group Sequence Policy Optimization (GSPO) with an LLM-as-Judge reward function.

Aryabhata 2 is a reasoning-focused language model developed by PhysicsWallah for competitive STEM examinations (JEE, NEET). It is obtained by post-training GPT-OSS-20B via reinforcement learning on a curated curriculum of Physics, Chemistry, Mathematics, and General Reasoning questions — achieving…

> VIDRAFT FINAL-Bench — chemistry-specialized 218B MoE, served via the DELPHI 5-Phase inference cascade.

Infers maternal and paternal transmitted and non-transmitted alleles from phased trio genotype data. The package supports SNP-level analyses of genetic nurture and transgenerational effects. It interoperates with Bioconductor VCF infrastructure through support for VariantAnnotation::VCF objects and returns R objects for downstream analysis.

A graph-based molecule modeling and chemoinformatics analysis toolkit fully implemented in Julia

The qmtools (quantitative metabolomics tools) package provides basic tools for processing quantitative metabolomics data with the standard SummarizedExperiment class. This includes functions for imputation, normalization, feature filtering, feature clustering, dimension-reduction, and visualization to help users prepare data for statistical analysis. This package also offers a convenient way to compute empirical Bayes statistics for which metabolic features are different between two sets of study samples. Several functions in this package could also be used in other types of omics data.

GENATATOR-PIPELINE is a Hugging Face pipeline for ab initio gene annotation from genomic DNA. It accepts a FASTA file, finds candidate transcript intervals, assigns transcript type, predicts exon and CDS structure, and writes a GFF3 annotation file.

Robert Johansson.

End-to-end composable multi-agent framework for automating OpenFOAM-based CFD simulations from natural language prompts, managing meshing, case setup, execution, error correction, and post-processing; achieves 100% success rate on 110 FoamBench tasks with Claude Opus 4.6 through Architect-Input Writer-Runner-Reviewer agent collaboration with RAG-enhanced generation and MCP tool integration (RPI CSML, 242+ stars, MIT License)

HealthJudge is a domain-adapted helpfulness evaluator for health-related Community Notes. It is designed to judge whether a note provides helpful context for a potentially misleading social-media post, following the Community Notes helpfulness criteria.

sosta (Spatial Omics STructure Analysis) is a package for analyzing spatial omics data to explore tissue organization at the anatomical structure level. It reconstructs anatomically relevant structures based on molecular features or cell types. It further calculates a range of metrics at the structure level to quantitatively describe tissue architecture. The package is designed to integrate with other packages for the analysis of spatial omics data.

Modular multi-agent scientific research assistant that automates idea generation, literature review, methodology design, code execution in Docker, visualization, LaTeX paper writing, and peer-review simulation across 10+ disciplines; winner of the NeurIPS 2025 Fair Universe Competition (573+ stars, GPL-3.0, 2025-2026)

gINTomics is an R package for Multi-Omics data integration and visualization. gINTomics is designed to detect the association between the expression of a target and of its regulators, taking into account also their genomics modifications such as Copy Number Variations (CNV) and methylation. What is more, gINTomics allows integration results visualization via a Shiny-based interactive app.

Makes alchemical free energy calculations easier by leveraging the full power and flexibility of the PyData stack.

## Model Description ProtGPT3-112M is a single-sequence autoregressive protein language model for protein sequence generation. It is the smallest model in the ProtGPT3 family, an open-source suite of promptable and aligned protein language models ranging from 112M to 10B parameters.

ProtGPT3-10B is a single-sequence autoregressive protein language model for protein sequence generation. It is the largest model in the ProtGPT3 family, an open-source suite of promptable and aligned protein language models ranging from 112M to 10B parameters.

CalibraCurve is a computational tool designed to generate calibration curves for targeted mass spectrometry-based quantitative data. It is applicable to various omics disciplines, including proteomics, lipidomics, and metabolomics. The package also offers functionalities for data and calibration curve visualization and concentration prediction from new datasets based on the established curves.

The package provides `rlang` data masks for the SummarizedExperiment class. The enables the evaluation of unquoted expression in different contexts of the SummarizedExperiment object with optional access to other contexts. The goal for `plyxp` is for evaluation to feel like a data.frame object without ever needing to unwind to a rectangular data.frame.

Programmatic framework for designing state-switching proteins via backpropagation through compositional design constraints parameterized by structure prediction models; enables de novo design of allosteric regulators and fluorescent biosensors for arbitrary small-molecule analytes (79+ stars, MIT License, ICML 2026)

Efficient differentiable n-dimensional PDE solvers built on JAX and Equinox, shipping 46+ built-in equations with Fourier spectral methods, exponential time differencing, and full auto-differentiation for physics-based deep learning workflows (MIT, 200+ stars, 2024)

An ultrafast and memory-efficient tool for aligning sequencing reads to long reference sequences.

PhyloProfile is a tool for exploring complex phylogenetic profiles. Phylogenetic profiles, presence/absence patterns of genes over a set of species, are commonly used to trace the functional and evolutionary history of genes across species and time. With PhyloProfile we can enrich regular phylogenetic profiles with further data like sequence/structure similarity, to make phylogenetic profiling more meaningful. Besides the interactive visualisation powered by R-Shiny, the package offers a set of further analysis features to gain insights like the gene age estimation or core gene identification.

This repository hosts release artifacts for ReCLIP:

A haplotype-resolved assembler for accurate Hifi reads.

The `TrIdent` R package automates the analysis of transductomics data by detecting, classifying, and characterizing read coverage patterns associated with potential transduction events. Transductomics is a DNA sequencing-based method for the detection and characterization of transduction events in pure cultures and complex communities. Transductomics relies on mapping sequencing reads from a viral-like particle (VLP)-fraction of a sample to contigs assembled from the metagenome (whole-community) of the same sample. Reads from bacterial DNA carried by VLPs will map back to the bacterial contigs of origin creating read coverage patterns indicative of ongoing transduction.

Interactive R package with an intuitive Shiny-based graphical interface for alternative splicing quantification and integrative analyses of alternative splicing and gene expression based on The Cancer Genome Atlas (TCGA), the Genotype-Tissue Expression project (GTEx), Sequence Read Archive (SRA) and user-provided data. The tool interactively performs survival, dimensionality reduction and median- and variance-based differential splicing and gene expression analyses that benefit from the incorporation of clinical and molecular sample-associated features (such as tumour stage or survival). Interactive visual access to genomic mapping and functional annotation of selected alternative splicing events is also included.

!Protein-ligand interaction header

GGUF quantizations of HuggingFaceBio/Carbon-3B — a generative DNA foundation model — for efficient inference with llama.cpp.

Banksy is an R package that incorporates spatial information to cluster cells in a feature space (e.g. gene expression). To incorporate spatial information, BANKSY computes the mean neighborhood expression and azimuthal Gabor filters that capture gene expression gradients. These features are combined with the cell's own expression to embed cells in a neighbor-augmented product space which can then be clustered, allowing for accurate and spatially-aware cell typing and tissue domain segmentation.

Extends beachmat to initialize tatami matrices from TileDB-backed arrays. This allows C++ code in downstream packages to directly call the TileDB C/C++ library to access array data, without the need for block processing via DelayedArray. Developers only need to import this package to automatically extend the capabilities of beachmat::initializeCpp to TileDBArray instances.

Tool for converting raw DNA data files between 23andMe, AncestryDNA, MyHeritage, and FamilyTreeDNA formats.

High-throughput PubChem client for batch queries with caching, validation, rate-limit-aware retries, and a simple CLI.

A HuggingFace-compatible repackaging of PlasmidGPT (Shao, 2024) — a GPT-2-style decoder pretrained on 153k engineered plasmid sequences from Addgene. Loadable with standard AutoModelForCausalLM and AutoTokenizer. Used as the base for PlasmidGPT-SFT and PlasmidGPT-GRPO.

A Flexible Model For Record Linkage

First fully open-source model achieving AlphaFold3-level accuracy with 1000x faster binding affinity prediction (MIT)

1. 概述 2. 数据处理流程 3. 模型架构 4. 损失函数设计 5. 训练流程 6. 配置参数

Hulu-Med: A Transparent Generalist Model towards Holistic Medical Vision-Language Understanding

Hulu-Med: A Transparent Generalist Model towards Holistic Medical Vision-Language Understanding

Hulu-Med: A Transparent Generalist Model towards Holistic Medical Vision-Language Understanding

Highly scalable equivariant deep learning interatomic potentials enabling million-atom molecular dynamics simulations with ab initio accuracy, building on E(3)-equivariant architectures for large-scale atomistic modeling (mir-group, MIT License, 480+ stars)

First architecture deeply integrating a DNA foundation model with an LLM for multimodal biological reasoning, achieving 98% accuracy on KEGG disease pathway prediction and 15%+ average gains on variant effect prediction with interpretable step-by-step reasoning traces (bowang-lab, 390+ stars)

Foundational auto-research agent framework for agentic science at scale, providing modular agent construction, run-level self-evolution, and multiple SciMaster domain agents (ML-Master, X-Master, Browse-Master); outperforms general-purpose agents across authoritative benchmarks including the OpenAI Frontier Science Benchmark (206+ stars, Apache 2.0, 2026)

SSSOM is a Simple Standard for Sharing Ontological Mappings, providing - a TSV-based representation for ontology term mappings - a comprehensive set of standard metadata elements to describe mappings and - a standard translation between the TSV and the Web Ontology Language (OWL). Most metadata elements, such as "sssom:mapping_justification" are defined in the sssom namespace.

MS-based metabolomics data processing and compound annotation pipeline.

Performs feature annotations on LC-MS All-ion fragmentation datasets using fragment ion libraries.

Decentralized self-organizing teams of AI agents for long-running computational scientific experimentation; agents critique each other's proposals before spending compute and share successes/failures to avoid redundant exploration, achieving +8.33% on BioML-Bench, 1.9× faster nanoGPT optimization, and +12.5% on ProteinGym ACE2-Spike (425+ stars, 2026)

Generative foundation model for functional antibody and nanobody design, supporting de novo generation, affinity maturation, inverse design, structure prediction, and humanization (Tencent AI4S, ICLR 2025)

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

With the dedicated fortify method implemented for flowSet, ncdfFlowSet and GatingSet classes, both raw and gated flow cytometry data can be plotted directly with ggplot. ggcyto wrapper and some customed layers also make it easy to add gates and population statistics to the plot.