Find open-source science resources

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)

# Overview This is the CellHermes model, based on the LLaMA-3.1-8B-instruct architecture developed by Meta, fine-tuned using single-cell RNA sequencing (scRNA-seq) datasets from CellxGene and PPI network from BioGRID. CellHermes is an innovative framework for adapting existing large language models…

Self-hostable scientific claim-verification and literature-review tool combining Semantic Scholar retrieval, bibliometric scoring, and LLM-based evidence synthesis for large-batch validation workflows

Learnable latent embeddings for joint behavioral and neural analysis, enabling consistent and interpretable mapping of neural activity to behavior across modalities, species, and experiments (EPFL & Harvard, 1K+ stars)

# or·a·cle /ˈôrəkəl/ — a source of wise counsel; one who provides authoritative knowledge. From Latin ōrāculum, meaning divine announcement. In computer science, an oracle is a black box that always returns the correct answer — you don't ask it how it knows, you ask and it answers.

This repository contains Chemistry fine-tuned Qwen3-4B checkpoints from the local SciKnowEval-style generalization setup.

PlantGeneAnn is a plant genome foundation model that enables the prediction of various plant genomic elements at single-nucleotide resolution. The model is built upon the PlantBiMoE architecture with a 1D U-Net segmentation head, specifically designed for automated plant genome annotation.

👋 Join our LiGHT community. 📖 Check out the MeditronFO blog and MeditronFO preprint. 🔜 If you are a clinician join the MOOVE initiative here.

Sexo-FR is a French-language conversational language model that provides reliable, caring, and evidence-based sexual health information (information en santé sexuelle). It is part of a French public-health initiative whose goal is to make trustworthy sexual-health information more accessible to the…

👋 Join our LiGHT community. 📖 Check out the MeditronFO blog and MeditronFO preprint. 🔜 If you are a clinician join the MOOVE initiative here.

Gravity-bio-16B-A3B is a biology-focus midtrained model derived from Gravity-16B-A3B-Base. It uses the same sparse Mixture-of-Experts (MoE) architecture and tokenizer as Gravity-16B-A3B-Base, with additional midtraining for biological understanding on TheBioCollection corpus.

Automated academic illustration generation for AI scientists, converting research papers into publication-ready figures using VLMs and diffusion models with iterative refinement (PKU & Google Research, 6.2K+ stars, 2026)

Jolia is a 3D CT foundation model that encodes images into vector representations program. It encodes a whole 3D CT volume into:

First multi-domain generative foundation model for the natural sciences built on a unified scientific grammar, encoding proteins, antibodies, small molecules, chemical reactions, materials, and their spatial interactions into a shared token vocabulary; enables unified generation, prediction, and design across domains under a purely autoregressive paradigm (134+ stars, Apache 2.0, 2026)

Molexar-10M Base is the unconditional base model for Molexar, a unified multimodal molecular foundation model for drug design. It is trained as an autoregressive molecular language model over Fragment-SELFIES, a BRICS-fragment molecular language with validity-preserving decoding and…

Molexar-10M Omni is the universal multi-condition model for Molexar, a unified multimodal molecular foundation model for drug design. It starts from fairydance/molexar-10m-base and is supervised fine-tuned to generate Fragment-SELFIES molecules under scalar molecular-property,…

GRamma-12B is a 12-billion-parameter instruction-tuned language model specialized for the Greek medical domain. It is built on top of Gemma 3 12B Instruct and adapted through parameter-efficient fine-tuning on a collection of Greek and bilingual medical question-answering data.

A powerful, high-performance bioinformatics framework for discovering, evaluating, and verifying microRNA (miRNA) interactions across DNA, RNA, and protein target sequences. The mirnaprotpred package provides two core modules: SeqFinder: A discovery engine to find all potential miRNA interactions across a genome or target sequence. Validator: A targeted verification engine to test specific, user-provided miRNAs against a target sequence. Both modules are powered by a shared, rigorous biological engine that evaluates exact seed matching, wobble pairing, AU-rich context, and RNAduplex thermodynamic stability.

Full weight-level fine-tuning of InstaDeepAI/nucleotide-transformer-v2-50m-multi-species for binary DNA sequence classification on two GenomicBenchmarks tasks. All parameters are updated rather than using LoRA or a frozen backbone, with a leakage-free train/validation/test protocol and multi-seed…

Cross-domain foundation model for continuum dynamics trained on 19 physical scenarios spanning 63 variables, featuring adaptive compute via stride modulation and patch jittering for long-run stability (Polymathic AI, 293+ stars, MIT License)

QLoRA adapter for Llama-3.1-8B-Instruct, fine-tuned on PubMedQA for yes / no / maybe biomedical question answering (run5).

BioMatrix is a multimodal biological foundation model that natively integrates 1D sequences, 3D structures, and natural language for both molecules and proteins within a single decoder-only architecture.

Autonomous multi-agent AI scientist that mirrors a PhD workflow: literature review → grounded hypothesis → real ML experiments → self-critique → write-up; features a deterministic harness with frozen-metric scoring, edit allowlists, and a verified registry to make reward-hacking and hallucination impossible, plus 108 unit tests runnable without API keys or GPUs (461+ stars, MIT License, 2026)

A domain-adapted clinical LLM fine-tuned on synthetic Indian medical Q&A records using QLoRA (4-bit quantization) with Unsloth 2x speedup. Built to power the conversational AI layer.

Foundation model for universal prompt-driven medical image segmentation extending SAM3 to clinical imaging, supporting 2D public benchmarks and 3D training/evaluation with text and box prompts; pretrained weights available on HuggingFace (189+ stars)

Numerical differential equation solving in JAX

ProtGPT3-MSA is a multiple-sequence, homolog-conditioned autoregressive protein language model. It is part of the ProtGPT3 family, an open-source suite of promptable and aligned protein language models for protein sequence generation.

KAU-BioMedLLM is a research prototype for source-grounded biomedical variant interpretation. The current public release contains the LoRA adapter and documentation for a guarded report-generation system built around a curated biomedical evidence panel, citation enforcement, and abstention when…

ApeTokenizer-SMILES is an Atom Pair Encoding (APE) tokenizer for SMILES strings, trained on ~2M unique canonical SMILES from ChEMBL 36. It is the SMILES counterpart to ApeTokenizer-SELFIES, released alongside ModernMolBERT.

This model is an NF4 (Normal Float 4-bit) quantized version of the base model InstaDeepAI/nucleotide-transformer-2.5b-multi-species. The checkpoint was quantized using the BitsAndBytes library with double quantization enabled and BF16 computation.

Fudan University's cascade machine learning forecasting system for 15-day global weather prediction, employing a 3D Earth-specific transformer with hard-constraint techniques to achieve state-of-the-art accuracy against traditional NWP and AI baselines

Diffusion-based document OCR framework replacing autoregressive decoding with block-level parallel diffusion decoding, enabling high-accuracy text recognition in scientific PDFs (613+ stars, MIT License)

## Important Notice If you are using GENERator for sequence generation, please ensure that the length of each input sequence is a multiple of 6. This can be achieved by either: 1. Padding the sequence on the left with 'A' (left padding); 2. Truncating the sequence from the left (left truncation).

Cross-platform system optimizations for accelerating AlphaFold3 training with 1.73x speedup and 1.23x memory reduction

Learning the language of protein-protein interactions

Genomic foundation model for metagenomic and genome annotation, featuring an 8k base-pair context and 500M parameters trained on 386B base pairs of eukaryotic DNA; provides expert models and a unified CLI for prokaryotic/eukaryotic coding-sequence annotation with strong performance on Genomic Benchmarks, Nucleotide Transformer tasks, and custom Gener tasks (GenerTeam, 314+ stars, MIT License)

Automatic atomic model building program for cryo-EM maps using deep learning, enabling rapid de novo protein structure determination from electron density with high accuracy (3DEM/EMBL, 169+ stars)

Performs laboratory unit conversions across molarity, OD600 cell density, C₁V₁ dilution, and related dimensional pairs from mass, volume, molecular weight, and organism-specific OD factors. A browser calculator combines four modes in one tabbed workspace with compound MW lookup, species-aware OD uncertainty ranges, cross-tab chaining, and shareable links; a Python library and command-line tool submit the same parameters to the Pepkio Tools API for scripted use. Calculator arithmetic for the API client is hosted remotely; the client transmits conversion inputs and returns structured results and shareable run identifiers.

Translates spectrophotometer and NanoDrop readings into mass and molar concentrations for dsDNA, ssDNA, ssRNA, and protein from a single anchor input, with optional sequence-specific nearest-neighbor extinction coefficients. A browser calculator supports bidirectional unit conversion, batch processing of up to ninety-six NanoDrop export rows, and A260/A280 purity interpretation with plain-language warnings; a REST API exposes converter, batch, and purity modes for scripted use. Calculator arithmetic is hosted remotely; API clients transmit parameters and return structured result fields and shareable run identifiers.

Calculates sequence-derived molecular properties and related laboratory planning outputs from FASTA and assay setup inputs. The tool supports sequence analysis for DNA, RNA, and protein entries, plus dilution and ligation calculation modes through one API-backed workflow. Programmatic use is available through a Python library and command-line interface that submit run payloads and return structured result objects.

Translates between centrifuge RPM and relative centrifugal force using rotor geometry, reporting g-force or speed at rmin, ravg, and rmax. Convert mode handles rpm_to_rcf and rcf_to_rpm with rotor presets or manual radii in mm; transfer mode maps a source RPM on one rotor to an equivalent target RPM at matched rmax RCF; batch mode processes multiple spin steps from CSV or row arrays. A browser calculator and a Python library with command-line interface submit the same parameters to the Pepkio Tools API and return structured results with optional methods text and safety warnings.

Derives cells per well and suspension pipette volumes for standard 6-, 12-, 24-, 48-, 96-, and 384-well plates from a hemocytometer stock count, trypan blue viability, and target seeding confluency, with QC flags for low viability and impractical transfers. A browser calculator supports interactive planning with cell-line presets; a Python library and command-line tool submit the same parameters to the Pepkio Tools API for scripted and pipeline use. Calculator arithmetic is hosted remotely; the client transmits parameters and returns structured plate tables and shareable run identifiers.

Constructs Punnett squares and offspring genotype and phenotype ratios for complete, incomplete, codominant, ABO multiple-allele, and sex-linked Mendelian crosses from parent genotypes, with step-by-step walkthroughs and reduced ratio output. A browser calculator provides live grids, textbook presets, PNG and SVG export, and shareable links; a Python library and command-line tool submit the same parameters to the Pepkio Tools API for scripted use. Calculator arithmetic for the API client is hosted remotely; the client transmits cross inputs and returns structured grids, ratios, walkthroughs, and shareable run identifiers.

Evaluates Hardy-Weinberg equilibrium for diploid loci with two to six alleles using chi-square and Guo-Thompson exact tests, inbreeding coefficient F, and plain-language verdicts from observed genotype counts, allele frequencies, or biallelic disease incidence. A browser calculator provides De Finetti plots, export, and Wright-Fisher simulation under selection, drift, mutation, and migration; a Python library and command-line tool submit the same parameters to the Pepkio Tools API for scripted use. Calculator arithmetic for the API client is hosted remotely; the client transmits genotype or simulation inputs and returns structured results and shareable run identifiers.

Performs batch four-parameter and five-parameter logistic regression on multi-compound concentration–response screens to estimate IC50, EC50, pIC50, Hill slope, and related potency metrics with per-compound QC grades. A browser calculator supports CSV upload, curve review, and figure export; a Python library and command-line tool submit the same parameters to the Pepkio Tools API for scripted and pipeline use. Calculator arithmetic is hosted remotely; the client transmits concentration–response data and returns structured fit results and shareable run identifiers.

Processes 96-well plate absorbance data through blank subtraction, regression fitting, and dilution correction to report sample concentrations with QC flags for BCA, Bradford, and ELISA workflows. A browser calculator supports interactive grid entry with CSV and PDF export; a Python library and command-line tool submit the same parameters to the Pepkio Tools API for scripted and pipeline use. Calculator arithmetic is hosted remotely; the client transmits plate layout and absorbance values and returns model comparison, per-sample concentrations, and shareable run identifiers.

Medical-GPT-OSS-Swallow-120B is a medical-domain language model based on tokyotech-llm/GPT-OSS-Swallow-120B-RL-v0.1. It is designed to support research and development toward safe and trustworthy AI for Japanese clinical settings.