Find open-source science resources

Dynamic Protein Data Bank integrating dynamic behaviors and physical properties into protein structures via a new dataset and SE(3) model extension, enabling richer understanding of protein conformational landscapes (Fudan University, 784+ stars)

# LAMAR LAMAR is a Foundation Language Model for RNA Regulation, which achieves better or comparable performance compared to baseline models in various RNA regulation tasks, helping to decipher the rules of RNA regulation.

methylscaper is an R package for processing and visualizing data jointly profiling methylation and chromatin accessibility (MAPit, NOMe-seq, scNMT-seq, nanoNOMe, etc.). The package supports both single-cell and single-molecule data, and a common interface for jointly visualizing both data types through the generation of ordered representational methylation-state matrices. The Shiny app allows for an interactive seriation process of refinement and re-weighting that optimally orders the cells or DNA molecules to discover methylation patterns and nucleosome positioning.

dandelionR is an R package for performing single-cell immune repertoire trajectory analysis, based on the original python implementation. It provides the necessary functions to interface with scRepertoire and a custom implementation of an absorbing Markov chain for pseudotime inference, inspired by the Palantir Python package.

Specialized model for Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature

Specialized model for Chemical Entity Recognition - Identifies chemical compounds and substances in biomedical literature

Specialized model for Chemical Entity Recognition - Chemical entities from the BC5CDR dataset

A specialized medical embedding model fine-tuned from Clinical ModernBERT using InfoNCE contrastive learning on PubMed title-abstract pairs.

Deep Graph Library for scalable deep learning on graphs, powering molecular modeling, materials discovery, protein interaction networks, and scientific knowledge graph learning across PyTorch, TensorFlow, and MXNet backends (14K+ stars)

Scientific equation discovery and symbolic regression using LLMs, combining code generation with evolutionary search (ICLR 2025 Oral)

Scafari is a Shiny application designed for the analysis of single-cell DNA sequencing (scDNA-seq) data provided in .h5 file format. The analysis process is structured into the four key steps "Sequencing", "Panel", "Variants", and "Explore Variants". It supports various analyses and visualizations.

ChromSCape - Chromatin landscape profiling for Single Cells - is a ready-to-launch user-friendly Shiny Application for the analysis of single-cell epigenomics datasets (scChIP-seq, scATAC-seq, scCUT&Tag, ...) from aligned data to differential analysis & gene set enrichment analysis. It is highly interactive, enables users to save their analysis and covers a wide range of analytical steps: QC, preprocessing, filtering, batch correction, dimensionality reduction, vizualisation, clustering, differential analysis and gene set analysis.

AI agent for therapeutic reasoning across a universe of tools, achieving 92.1% accuracy in drug reasoning and outperforming GPT-4o by 25.8% (Harvard MIMS, 2025)

Multi-modal geospatial ML platform for agriculture and sustainability, fusing satellite imagery (RGB, SAR, multispectral), drone imagery, weather data, and sensor data for crop identification, carbon footprint estimation, and microclimate prediction (Microsoft Research, MIT License)

Family of diffusion protein language models demonstrating versatile generative and predictive capabilities for protein sequences and structures, including multimodal co-generation, conditional folding, inverse folding, motif scaffolding, and representation learning, with open pretrained weights and training scripts (327+ stars, ICML 2024, ICLR 2025, ICML 2025 Spotlight)

JAMUN is a novel approach for generating conformational ensembles of protein structures, presented in the paper JAMUN: Bridging Smoothed Molecular Dynamics and Score-Based Learning for Conformational Ensembles.

Large-scale flow-based protein backbone generator utilizing hierarchical fold class labels for conditioning with a tailored scalable transformer architecture, enabling controllable de novo protein design (264+ stars)

Image segmentation is the process of identifying the borders of individual objects (in this case cells) within an image. This allows for the features of cells such as marker expression and morphology to be extracted, stored and analysed. simpleSeg provides functionality for user friendly, watershed based segmentation on multiplexed cellular images in R based on the intensity of user specified protein marker channels. simpleSeg can also be used for the normalization of single cell data obtained from multiple images.

Many tools for data analysis are not available in R, but are present in public repositories like conda. The Herper package provides a comprehensive set of functions to interact with the conda package managament system. With Herper users can install, manage and run conda packages from the comfort of their R session. Herper also provides an ad-hoc approach to handling external system requirements for R packages. For people developing packages with python conda dependencies we recommend using basilisk (https://bioconductor.org/packages/release/bioc/html/basilisk.html) to internally support these system requirments pre-hoc.

syntenet can be used to infer synteny networks from whole-genome protein sequences and analyze them. Anchor pairs are detected with the MCScanX algorithm, which was ported to this package with the Rcpp framework for R and C++ integration. Anchor pairs from synteny analyses are treated as an undirected unweighted graph (i.e., a synteny network), and users can perform: i. network clustering; ii. phylogenomic profiling (by identifying which species contain which clusters) and; iii. microsynteny-based phylogeny reconstruction with maximum likelihood.

Provides a comprehensive framework for representing, analyzing, and visualizing genomic interactions, particularly focusing on gene-enhancer relationships. The package extends the GenomicRanges infrastructure to handle paired genomic regions with specialized methods for chromatin interaction data from Hi-C, Promoter Capture Hi-C (PCHi-C), and single-cell ATAC-seq experiments. Key features include conversion from common interaction formats, annotation of promoters and enhancers, distance-based analyses, interaction strength metrics, statistical modeling using CHiCANE methodology, and tailored visualization tools. The package aims to standardize the representation of genomic interaction data while providing domain-specific functions not available in general genomic interaction packages.

MOGAMUN is a multi-objective genetic algorithm that identifies active modules in a multiplex biological network. This allows analyzing different biological networks at the same time. MOGAMUN is based on NSGA-II (Non-Dominated Sorting Genetic Algorithm, version II), which we adapted to work on networks.

piRNAs (short for PIWI-interacting RNAs) and their PIWI protein partners play a key role in fertility and maintaining genome integrity by restricting mobile genetic elements (transposons) in germ cells. piRNAs originate from genomic regions known as piRNA clusters. The piRNA Cluster Builder (PICB) is a versatile toolkit designed to identify genomic regions with a high density of piRNAs. It constructs piRNA clusters through a stepwise integration of unique and multimapping piRNAs and offers wide-ranging parameter settings, supported by an optimization function that allows users to test different parameter combinations to tailor the analysis to their specific piRNA system. The output includes extensive metadata columns, enabling researchers to rank clusters and extract cluster characteristics.

Coralysis is an R package featuring a multi-level integration algorithm for sensitive integration, reference-mapping, and cell-state identification in single-cell data. The multi-level integration algorithm is inspired by the process of assembling a puzzle - where one begins by grouping pieces based on low-to high-level features, such as color and shading, before looking into shape and patterns. This approach progressively blends the batch effects and separates cell types across multiple rounds of divisive clustering.

Highly focused on medical Training datasets ; + Upgraded inplace

topGO package provides tools for testing GO terms while accounting for the topology of the GO graph. Different test statistics and different methods for eliminating local similarities and dependencies between GO terms can be implemented and applied.

DeepSeek's open-source large language model for formal theorem proving in Lean 4, integrating informal and formal mathematical reasoning through recursive subgoal decomposition and reinforcement learning powered by DeepSeek-V3, with open weights and ProverBench evaluation (2025)

S4 generic functions and classes needed by Bioconductor proteomics packages.

This model classifies facial skin images into 6 common dermatological conditions using a fine-tuned EfficientNetV2B0 architecture.

Easily get SRA download links and other information.

darkknight25/deepseek-16b-medical-GPT is a fine-tuned version of deepseek-ai/deepseek-l6b-moe-chat, optimized for medical question answering, reasoning, and clinical summarization using QLoRA and open-access healthcare datasets.

Unsloth Dynamic 2.0 achieves superior accuracy & outperforms other leading quants.

This is a merge of pre-trained language models created using mergekit, combining the specialty and general reasoning skills of Esper 3 8b and Shining Valiant 3 8b.

Most human genes have multiple promoters that control the expression of different isoforms. The use of these alternative promoters enables the regulation of isoform expression pre-transcriptionally. Alternative promoters have been found to be important in a wide number of cell types and diseases. proActiv is an R package that enables the analysis of promoters from RNA-seq data. proActiv uses aligned reads as input, and generates counts and normalized promoter activity estimates for each annotated promoter. In particular, proActiv accepts junction files from TopHat2 or STAR or BAM files as inputs. These estimates can then be used to identify which promoter is active, which promoter is inactive, and which promoters change their activity across conditions. proActiv also allows visualization of promoter activity across conditions.

High-throughput single-cell measurements of DNA methylation allows studying inter-cellular epigenetic heterogeneity, but this task faces the challenges of sparsity and noise. We present vmrseq, a statistical method that overcomes these challenges and identifies variably methylated regions accurately and robustly.

Therapeutics Data Commons: 66 AI-ready datasets across 22 drug discovery tasks with 29 leaderboards, covering target identification, molecular generation, ADMET prediction, and clinical trial outcomes (Harvard MIMS, NeurIPS 2021/2024)

A library for estimating thermochemical properties of molecules and adsorbates using group additivity.

Using llama.cpp release b5868 for quantization.

Segment Anything in 3D medical images and videos, extending SAM2 to volumetric and temporal medical imaging with state-of-the-art zero-shot segmentation performance across CT, MRI, and surgical video (arXiv 2025)

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

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

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

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

Segment Anything in 3D Medical Images and Videos

> A CMR-report contrastive model combining Vision Transformers and pretrained text encoders.

Unsloth Dynamic 2.0 achieves superior accuracy & outperforms other leading quants.