Find open-source science resources

Benchmark evaluating AI agents' ability to replicate 20 ICML 2024 Spotlight/Oral papers from scratch, with 8,316 gradable tasks and author-co-developed rubrics

Pretrained time series foundation model for zero-shot forecasting across diverse scientific and real-world domains; tokenizes continuous time series into discrete bins to train transformer language models on large-scale corpora, achieving strong zero-shot generalization and competitive performance with task-specific supervised models on climate, energy, and health benchmarks (5.3K+ stars, Apache 2.0, 2024-2026)

This repository contains the model used for the paper Bridging Quantum Mechanics to Organic Liquid Properties via a Universal Force Field。

Integrates various levels of epigenomic information, including ChIP-seq, histone modification, ATAC-seq, and RNA-seq data. Regulatory network analysis uses combinatory approaches to infer regions of significance, such as enhancers. Downstream analysis identifies co-occurrence of epigenomic data at regions of interest. Visualization functions display multi-track genomic views with signal overlays. Please contact <ammawla@ucdavis.edu> for suggestions, feedback, or bug reporting.

R package for analysis of transcript and translation features through manipulation of sequence data and NGS data like Ribo-Seq, RNA-Seq, TCP-Seq and CAGE. It is generalized in the sense that any transcript region can be analysed, as the name hints to it was made with investigation of ribosomal patterns over Open Reading Frames (ORFs) as it's primary use case. ORFik is extremely fast through use of C++, data.table and GenomicRanges. Package allows to reassign starts of the transcripts with the use of CAGE-Seq data, automatic shifting of RiboSeq reads, finding of Open Reading Frames for whole genomes and much more.

A domain-optimized reasoning model built on DeepSeek-R1-Distill-Qwen-32B, refined through a multi-stage pipeline of GPTQ quantization-aware training and QLoRA fine-tuning. Achieves 84% on MedQA — within 4 points of GPT-4o — in a ~20GB package that fits on a single L40/L40s GPU.

# ModernGENA base ModernGENA is a DNA foundation model based on ModernBERT (a modernized BERT-style encoder architecture) adapted for genomic sequence modeling. ModernGENA base is the 377M-parameter version introduced in the paper Back to BERT in 2026: ModernGENA as a Strong, Efficient Baseline for…

An extended version of SCimilarity, a metric-learning model for single-cell RNA-seq that maps cells to a unified 128-dimensional embedding space. The original model and method are described in:

MotifPeeker is used to compare and analyse datasets from epigenomic profiling methods with motif enrichment as the key benchmark. The package outputs an HTML report consisting of three sections: (1. General Metrics) Overview of peaks-related general metrics for the datasets (FRiP scores, peak widths and motif-summit distances). (2. Known Motif Enrichment Analysis) Statistics for the frequency of user-provided motifs enriched in the datasets. (3. Motif Discovery Enrichment Analysis) Statistics for the frequency of ab-initio discovered motifs enriched in the datasets and compared with known motifs.

Access to Biological Web Services from Python.

Descriptor library containing a variety of fingerprinting techniques, including the Smooth Overlap of Atomic Positions (SOAP).

Battlefield is a Swiss-army toolkit originally developed to define and extract spatial spots from specific tissue regions—such as front regions, niche borders, invasive margins, and cluster interfaces—using spatial transcriptomics data or clustered tissue maps. It has since been extended to support trajectory selection and layer inspection, and now provides a collection of low-level utilities for spatial transcriptomics analysis. These utilities are primarily intended to be reused within higher-level analytical packages. It is designed to work with sequencing-based platforms such as Visium at several resolutions and Visium HD(binned).

A list of pipeline resources.

The VISTA (Visualization and Integrated System for Transcriptomic Analysis) platform streamlines differential expression workflows by wrapping DESeq2 and edgeR into a SummarizedExperiment-based container with consistent metadata. The package includes visualization utilities, MSigDB enrichment helpers, and optional deconvolution support to simplify interactive exploration of RNA-seq experiments.

A python-based workflow manager.

Duchifat-2.3-Instruct is a state-of-the-art, instruction-tuned Large Language Model developed by TopAI. As the flagship of the Duchifat series, this model represents a fundamental breakthrough in how Hebrew is processed, reasoned, and generated in the LLM era.

Determine sample ploidy via flow cytometry histogram analysis. Reads Flow Cytometry Standard (FCS) files via the flowCore bioconductor package, and provides functions for determining the DNA ploidy of samples based on internal standards.

!image

Computational toolbox for large scale Calcium Imaging Analysis, including movie handling, motion correction, source extraction, spike deconvolution and result visualization, using machine learning for automated neuron detection and activity inference in two-photon and one-photon calcium imaging data (723+ stars, actively maintained)

An extensive set of data (pre-)processing and analysis methods and tools for metabolomics and other omics, with a strong emphasis on statistics and machine learning. This toolbox allows the user to build extensive and standardised workflows for data analysis. The methods and tools have been implemented using class-based templates provided by the struct (Statistics in R Using Class-based Templates) package. The toolbox includes pre-processing methods (e.g. signal drift and batch correction, normalisation, missing value imputation and scaling), univariate (e.g. ttest, various forms of ANOVA, Kruskal–Wallis test and more) and multivariate statistical methods (e.g. PCA and PLS, including cross-validation and permutation testing) as well as machine learning methods (e.g. Support Vector Machines). Ontology terms have been integrated to provide standardised definitions for the different methods, inputs and outputs.

Dataset and benchmarking framework integrating histology and spatial transcriptomics, enabling multimodal analysis of whole-slide images with matched spatial gene expression for advancing computational pathology and tissue microenvironment research (Mahmood Lab, Harvard Medical School, 411+ stars)

Create and plot diffusion maps.

Imports transcript-level abundance, estimated counts and transcript lengths, and summarizes into matrices for use with downstream gene-level analysis packages. Average transcript length, weighted by sample-specific transcript abundance estimates, is provided as a matrix which can be used as an offset for different expression of gene-level counts.

lisaClust provides a series of functions to identify and visualise regions of tissue where spatial associations between cell-types is similar. This package can be used to provide a high-level summary of cell-type colocalization in multiplexed imaging data that has been segmented at a single-cell resolution.

This package provides functions to browse the harmonized metadata for large omics databases. This package also supports data navigation if the metadata incorporates ontology.

Learning the language of protein-protein interactions

Fully autonomous medical image segmentation research system that generates complete manuscripts end-to-end from datasets with zero human intervention, beating strongest baselines on 24 of 31 datasets and achieving T1-T2 tier manuscript quality in double-blind evaluations (USTC & Shanghai AI Lab, 2026)

Provides hurdle negative binomial models for differential expression analysis with long-read RNA-Seq data.

First agentic LLM for autonomous data science with end-to-end pipeline from data to analyst-grade reports

Fine-tuned version of google/gemma-4-E4B-it across three professional domains — Medical, Legal, and Finance — using QLoRA (4-bit NF4) with Optuna-tuned hyperparameters, trained on Kaggle T4 GPU.

Scalable agentic training environment for code-centric reasoning in biomedical data science

This is an R package for interfacing with the BIOM file format. This package includes basic tools for reading biom-format files, accessing and subsetting data tables from a biom object (which is more complex than a single table), as well as limited support for writing a biom-object back to a biom-format file. The design of this API is intended to match the python API and other tools included with the biom-format project, but with a decidedly "R flavor" that should be familiar to R users. This includes S4 classes and methods, as well as extensions of common core functions/methods.

Stamps Seurat, SingleCellExperiment, and SummarizedExperiment objects with a persistent metadata passport. For Seurat objects the passport is stored in the misc slot; for SingleCellExperiment and SummarizedExperiment objects it is stored in the metadata slot. Tracks animal info, experiment details, lineage (parent/child relationships), RDS registry numbers, processing logs, and custom fields. Includes an interactive Shiny gadget to fill and update the passport, and a read mode to print the full passport to console. The passport persists inside the RDS file with no external files needed.

Multi-modal foundation model for biomolecular structure prediction (proteins, small molecules, DNA, RNA, glycans) achieving SOTA across benchmarks, with optional MSA/template support (Chai Discovery, 2024)

Open Parser for Systematic IUPAC nomenclature

The Context and Measurement Ontology (COMO) contains ontological terms to describe the context for various types of experimental data and measurements. It is useful in its current state for several different environmental microbiology projects. This ontology is used in multiple CORAL (Contextual Ontology-based Repository Analysis Library) deployments.

Toolbox for larger-than-memory scientific computing and visualization, providing efficient out-of-core data structures using files or shared memory, for dense and sparse vectors, matrices, and arrays, with applications to nonuniformly sampled signals and images.

Programmatic data labeling and weak supervision

Implements a DelayedArray backend for reading and writing dense or sparse arrays in the TileDB format. The resulting TileDBArrays are compatible with all Bioconductor pipelines that can accept DelayedArray instances.

Offline-first scientific writing workspace powered by Claude, integrating LaTeX, Python, and 100+ scientific skills with local execution, Zotero integration, and privacy-focused design (2026)

L1 (Learning Unit 1) is the first language model from Lunit and Lunit Consortium, purpose-built for the medical domain. Derived from Gravity-16B-A3B-Base, L1 is designed for clinical reasoning and decision support.

Provides a consistent C++ class interface for reading from a variety of commonly used matrix types. Ordinary matrices and several sparse/dense Matrix classes are directly supported, along with a subset of the delayed operations implemented in the DelayedArray package. All other matrix-like objects are supported by calling back into R.

Parameter/topology editor and molecular simulator with visualization capability.

Medical time series foundation model pretrained on 454B time points from heterogeneous clinical corpora spanning ICU physiological signals and hospital EHR, with continuous-time rotary positional encoding, frequency-specialized Mixture-of-Experts, and neural ODE extrapolation for zero-shot forecasting across irregular and multimodal temporal health data (Microsoft, 399+ stars, MIT License)

CodonTranslator is a protein-conditioned codon sequence generation model trained on the representative-only data_v3 release.

!Banner.

Lightweight Expression displaYer (plotter / viewer) of SummarizedExperiment object in R. This package provides a quick and easy Shiny-based GUI to empower a user to use a SummarizedExperiment object to view

Mutations that rapidly accumulate in viral genomes during a pandemic can be used to track the evolution of the virus and, accordingly, unravel the viral infection network. To this extent, sequencing samples of the virus can be employed to estimate models from genomic epidemiology and may serve, for instance, to estimate the proportion of undetected infected people by uncovering cryptic transmissions, as well as to predict likely trends in the number of infected, hospitalized, dead and recovered people. VERSO is an algorithmic framework that processes variants profiles from viral samples to produce phylogenetic models of viral evolution. The approach solves a Boolean Matrix Factorization problem with phylogenetic constraints, by maximizing a log-likelihood function. VERSO includes two separate and subsequent steps; in this package we provide an R implementation of VERSO STEP 1.