VPOD: Visual Physiology Opsin Database

Summary of VPOD v1.3

Contains 1,714 unique opsin genotypes and corresponding λ_max phenotypes collected across all animals from 120+ separate publications.
Curates all heterologously expressed, and a partial collection of physiologically inferred opsin genes.
Uses VPOD data and deepBreaks to show regression-based machine learning (ML) models reliably predict λ_max, account for epistatic (non-additive) effects, and identify functional amino acid sites.
Lays the groundwork for future robust exploration of molecular-evolutionary patterns governing phenotype.

Interactive access

VPOD Explorer

The VPOD Explorer is the interactive web interface for the database. It lets users search across records, apply advanced filters, visualize λ_max distributions, export queried results, and suggest new data for curator review.

Open the Explorer

Explorer vs. GitHub Database

VPOD on GitHub

Best for versioned releases, raw database files, formatted machine-learning subsets, notebooks, and fully reproducible local workflows.

VPOD Explorer

Best for browser-based querying, filtering, quick CSV export, citation access, and community suggestions for new data entries.

Database File Structure

VPOD provides both fully curated machine-learning ready datasets and the raw database files for custom querying using SQLite.

Formatted Data Subsets

Located in vpod_data/VPOD_1.3/formatted_data_subsets/. Subsets suitable for direct model training without requiring MySQL or sequence alignment.

xxx.txt - Unaligned data subsets.
xxx_aligned.txt - Aligned data subsets.
VPOD_xxx_het_1.3.fasta - Fully aligned and formatted subsets.
xxx_meta.tsv - Metadata files (species, accession, λ_max).

Raw Database Files

Located in vpod_data/VPOD_1.3/raw_database_files/. Load into SQLite to create custom datasets.

litsearch.csv - Literature search information.
references.csv - All publication references.
opsins.csv - Sequence data & taxonomic metadata.
heterologous.csv - Opsin phenotype data & experimental metadata.

Clone VPOD

git clone https://github.com/VisualPhysiologyDB/visual-physiology-opsin-db.git
cd visual-physiology-opsin-db
# Start exploring with vpod_main_wf.ipynb

Workflows & Analysis Notebooks

Primary ML Workflow

vpod_main_wf.ipynb is the primary notebook for users. It contains a full pipeline for creating a local instance of VPOD using SQLite, formatting datasets, and training ML models using deepBreaks.

Sequence Manipulation

Includes tools for generating chimeras, in-silico deep-mutational-scanning (DMS), and reciprocal mutagenesis to build theoretical opsin variants for model testing.

Phylogenetic Imputation

R-based tools (Phylogenetic_Imputation.Rmd) to load tree files, make λ_max predictions via phylogenetic imputation, and compare them directly against ML outputs.

Mine-n-Match (MNM)

Advanced workflow combining heterologous expression data with in-vivo correlations to augment datasets for more robust taxonomic subset modeling.

Looking to just run predictions? Check out the OPTICS Tool instead.

Citation & Credits

If you use OPTICS in your research, please cite:

Seth A. Frazer & Todd H. Oakley. Accessible and Robust Machine Learning Approaches to Improve the Opsin Genotype-Phenotype Map. Molecular Biology & Evolution, 2026.06.23,
https://doi.org/10.1093/molbev/msag138

Original VPOD Development:

Seth A. Frazer, Mahdi Baghbanzadeh, Ali Rahnavard, Keith A. Crandall, & Todd H Oakley. Discovering genotype-phenotype relationships with machine learning and the Visual Physiology Opsin Database (VPOD). GigaScience, 2024.09.01.
https://doi.org/10.1093/gigascience/giae073

The Back-Bone ML Pipeline for OPTICS: DeepBreaks

Mahdi Baghbanzadeh, Tyson Dawson, Bahar Sayoldin, Seth A. Frazer, Todd H. Oakley, Keith A. Crandall & Ali Rahnavard. deepBreaks identifies and prioritizes genotype–phenotype associations using machine learning. Scientific Reports, 2026.11.07.
https://doi.org/10.1038/s41598-025-25580-6