Find open-source science resources

KIMMDY is a biomolecular reaction emulator that simulates complex biochemical reactions across conformational ensembles using kinetic Monte Carlo. It leverages graph neural networks to model dynamic, large-scale systems involving successive, competing reactions in biomolecules like proteins and DNA.

scRevol is an RL-based model designed to infer tumor evolution and characterize clonal structure from single-cell RNA sequencing (scRNA-seq) data. It utilizes CNV profiles to reconstruct evolutionary trajectories and resolve subclonal diversity in cancer.

Viral Host UnveiLing Kit - A tool-kit for phage host prediction. This tool is for bacteriophage host prediction based on annotated genomic features and deep neural networks.

MedSlice is an open-source pipeline of fine-tuned open-source LLMs (best: Llama 3.1 8B) that automatically extracts History of Present Illness, Interval History, and Assessment and Plan sections from clinical progress notes, outperforming GPT-4o (F1 0.92) while keeping data local for privacy.

Provides an R wrapper for the implementation of FI-tSNE from the python package openTNSE. See Poličar et al. (2019) <doi:10.1101/731877> and the algorithm described by Linderman et al. (2018) <doi:10.1038/s41592-018-0308-4>.

An Open Source Unsupervised Algorithm for Discovery of Spontaneous Behaviors | Behavioral segmentation of open field in DeepLabCut, or B-SOID ("B-side"), is an unsupervised learning algorithm written in MATLAB that serves to discover behaviors that are not pre-defined by users | B-SOID: Behavioral segmentation of open field in DeepLabCut

Flame is an open source framework for model development, hosting, and usage in production environments. Flame is a web application, with a Python backend and a web-based graphic interface, which can be used as a desktop application or installed in a server receiving requests from multiple users. Models can be built starting from any collection of biologically annotated chemical structures, since the software supports structural normalization, molecular descriptor generation and machine learning building, using predefined workflows. The model building workflow can be customized from the graphic interface, selecting the type of normalization, molecular descriptors, and machine learning algorithm to be used from a panel of state-of-the-art methods implemented natively.

External validation of machine learning models with adaptive sample splitting

scDART (single cell Deep learning model for ATAC-Seq and RNA-Seq Trajectory integration) is a scalable deep learning framework that embed the two data modalities of single cells, scRNA-seq and scATAC-seq data, into a shared low-dimensional latent space while preserving cell trajectory structures. Furthermore, scDART learns a nonlinear function represented by a neural network encoding the cross-modality relationship simultaneously when learning the latent space representations of the integrated dataset.