Find open-source science resources

Methodology for supervised clustering of potentially many predictor variables, such as genes etc., in time series datasets Provides functions that help the user assigning genes to predefined set of model profiles.

Statistical analysis of peptide microarrays

Find the most characteristic gene ontology terms for groups of human genes. This package was created as a part of the thesis which was developed under the auspices of MI^2 Group (http://mi2.mini.pw.edu.pl/, https://github.com/geneticsMiNIng).

Horizon chart D3-based JavaScript library for DNA data.

CNVrd2 uses next-generation sequencing data to measure human gene copy number for multiple samples, indentify SNPs tagging copy number variants and detect copy number polymorphic genomic regions.

The fmcsR package introduces an efficient maximum common substructure (MCS) algorithms combined with a novel matching strategy that allows for atom and/or bond mismatches in the substructures shared among two small molecules. The resulting flexible MCSs (FMCSs) are often larger than strict MCSs, resulting in the identification of more common features in their source structures, as well as a higher sensitivity in finding compounds with weak structural similarities. The fmcsR package provides several utilities to use the FMCS algorithm for pairwise compound comparisons, structure similarity searching and clustering.

This package is intended to identify differentially expressed genes driven by Copy Number Alterations from samples with both gene expression and CNA data.

This package implements methods to calculate information accretion for a given version of the gene ontology and uses this data to calculate remaining uncertainty, misinformation, and semantic similarity for given sets of predicted annotations and true annotations from a protein function predictor.

Recurrent breakpoint gene detection on copy number aberration profiles.

This package facilitates reading, preprocessing and manipulating Codelink microarray data. The raw data must be exported as text file using the Codelink software.

rTRM identifies transcriptional regulatory modules (TRMs) from protein-protein interaction networks.

This package provides a web interface to compute transcriptional regulatory modules with rTRM.

D3 JavaScript based genome viewer. Constructs SVGs.

Automate common SAM & BAM conversions.

Molecular property prediction with unified API for diverse models and respresentations,

Deep learning system for de novo design of high-affinity protein binders, achieving strong binding across diverse target classes including challenging intracellular proteins with significantly higher success rates than traditional wet-lab screening methods (Google DeepMind, Nature 2024)

Multimodal generative AI assistant for computational pathology enabling interactive visual-language conversations over histopathology images for diagnostic reasoning, case discussion, and education, built on a Mistral-7B backbone with domain-specific fine-tuning (Mahmood Lab, Harvard Medical School, 1.2K+ stars)

Google DeepMind and Google Quantum AI's transformer-based neural-network decoder for quantum error correction, trained on real Sycamore quantum processor data to outperform tensor-network and correlated matching decoders at code distances 3 and 5, demonstrating ML's role in enabling fault-tolerant quantum computing (Nature 2024)

Ecological modeling and conservation AI

Mathematical reasoning

A Python program to compute quasi-harmonic thermochemical data from Gaussian frequency calculations.

The R package dmGsea provides efficient gene set enrichment analysis specifically for DNA methylation data. It addresses key biases, including probe dependency and varying probe numbers per gene. The package supports Illumina 450K, EPIC, and mouse methylation arrays. Users can also apply it to other omics data by supplying custom probe-to-gene mapping annotations. dmGsea is flexible, fast, and well-suited for large-scale epigenomic studies.

Tools for quanlity control, analysis and visulization of Illumina DNA methylation array data.

This software is meant to be used for classification of images of cell-based assays for neuronal surface autoantibody detection or similar techniques. It takes imaging files as input and creates a composite score from these, that for example can be used to classify samples as negative or positive for a certain antibody-specificity. The reason for its name is that I during its creation have thought about the individual picture as an archielago where we with different filters control the water level as well as ground characteristica, thereby finding islands of interest.

Provides an R wrapper for BWA alignment algorithms. Both BWA-backtrack and BWA-MEM are available. Convenience function to build a BWA index from a reference genome is also provided. Currently not supported for Windows machines.

Probabilistic analysis of probe reliability and differential gene expression on short oligonucleotide arrays.

This package is designed to model gene detection pattern of scRNA-seq through a binary factor analysis model. This model allows user to pass into a cell level covariate matrix X and gene level covariate matrix Q to account for nuisance variance(e.g batch effect), and it will output a low dimensional embedding matrix for downstream analysis.

Tools for compositional and other sample-level ecological analyses and visualizations tailored for single-cell RNA-seq data. SETA includes functions for taxonomizing celltypes, normalizing data, performing statistical tests, and visualizing results. Several tutorials are included to guide users and introduce them to key concepts. SETA is meant to teach users about statistical concepts underlying ecological analysis methods so they can apply them to their own single-cell data.

Membrane Protein-Lipid Interaction Database. A large-scale experimentally validated dataset of 80685 residue-level lipid contact annotations across 4712 membrane proteins derived from PDB crystal and cryo-EM structures. Provides pre-computed binary contact labels, continuous distance values, sequence-identity-based cluster assignments, and ready-made train-validation-test splits for machine learning.

Plant Compound Extractor is a desktop application that builds a ready-to-use, deduplicated library of 3D ligand structures for a given plant. It queries multiple natural-product and chemical databases (COCONUT, LOTUS, Wikidata, PubChem, PlantaeDB, USDA Dr. Duke's, KNApSAcK and IMPPAT) in parallel, resolves each compound against PubChem for a canonical structure, and falls back to direct source retrieval when needed. Retrieved structures are then converted to 3D using RDKit, with configurable conformer generation and physicochemical filters (molecular weight, rotatable bonds, ring size, etc.). It can also process a manually supplied compound list, or convert an existing folder of 2D structures to 3D.