About PHILHARMONIC DB
Purpose
Functional genomics remains heavily biased toward a handful of model organisms. Over 90% of experimentally validated protein functions derive from fewer than 10 species, yet the majority of biodiversity—including ecologically critical taxa (corals, symbiotic dinoflagellates), agriculturally important species, and human pathogens—lacks reliable functional annotation. PHILHARMONIC DB addresses this gap through "guilt-by-association" network inference, requiring only a sequenced proteome to produce functional predictions. This resource will enable: (i) hypothesis generation for unstudied organisms, (ii) comparative functional genomics across deep evolutionary timescales, (iii) identification of lineage-specific pathway innovations, and (iv) prioritization of genes for experimental follow-up in non-model systems. The dataset directly supports ongoing work in agricultural biology, coral reef biology, veterinary genomics, and pathogen biology.
This project is a collaboration of the Computation and Biology Group (PI: Bonnie Berger) at MIT, the Bioinformatics and Computational Biology Group (PI: Lenore Cowen) at Tufts, and the Machine Leanring Lab (PI: Adam Klivans) at UT Austin. Support is provided by NIH R35 MIRA grant 1R35GM141861, NSF grant 2531917, NSF GRFP 2141064, and the NSF AI institute for Foundations of Machine Learning for GPU support.
To browse the list of available species, click here.
How to use this database
This platform is designed to provide seamless web-based exploration of large-scale functionally-annotated clusters generated by the PHILHARMONIC pipeline. You can navigate through our data using the following data resolution:
1. Browse species
On the home page, you will find a catalog of all supported species indexed in our current data release. Click on any species card to open its dedicated cluster browser.
2. Species page
Once inside a species view, you can browse individual functional clusters through our "cluster of clusters" visualization, where each circle represents a cluster of proteins. Further, you can browse or filter the list of clusters by their functional annotation and imputed discription/summary. Finally, if you have a specific protein of interest, you can query its accession in the search bar at the top of each species page. If that protein is assigned to a cluster, it will be recovered by the search and you can go directly to its Protein page. The Species page also contains links to directly download the species-wide predicted PPI network, as well as a MEDFORD-format file summarizing the parameters and corresponding outputs of the PHILHARMONIC pipeline run. Finally, the species page also specifies the NCBI Genome accession used.
3. Species page
The cluster is the core output of the PHILHARMONIC pipeline. The cluster page displays key information about each cluster, including its overall annotation and summary, its size, and the number of predicted PPI edges. The PPI network of proteins in that cluster is also displayed, ad individual proteins are clickable. Finally, A list of all GO terms assigned to any proteins and a list of all proteins in the cluster are included.
4. Protein page
The protein pages lists the annotations for a single protein, including all GO terms and Pfam domains. This page also links out to the corresponding NCBI Protein page for that protein, as well as linking to the protein's assigned cluster.
5. Downloads
The predicted PPI network for a given species can be downloaded from its Species page. Further, the Download link at the top of the website goes to an FTP-style interface where all PPI and annotation files are available for download.
Contributing
We welcome contributions and feedback on PHILHARMONIC DB. If you have questions, notice dataset errors, or want to contribute new species, please reach out to us by email at philharmonic-db@mit.edu.
Citing PHILHARMONIC
If you use PHILHARMONIC, its network predictions, or functional clusters in your research, including through this database, please cite our RECOMB 2025 paper, available as a preprint on bioRxiv:
@InProceedings{sledzieski2025philharmonic,
author = {Sledzieski, Sam and Versavel, Charlotte and Singh, Rohit and Ocitti, Faith and Devkota, Kapil and Kumar, Lokender and Shpilker, Polina and Roger, Liza and Yang, Jinkyu and Lewinski, Nastassja and Putnam, Hollie and Berger, Bonnie and Klein-Seetharaman, Judith and Cowen, Lenore},
title = {Decoding the Functional Interactome of Non-model Organisms with PHILHARMONIC},
booktitle = {Research in Computational Molecular Biology},
pages = {268--272},
year = {2025},
editor = {Sankararaman, Sriram},
series = {Lecture Notes in Computer Science},
volume = {15647},
doi = {10.1007/978-3-031-90252-9_19}
}