Interactive exploration of the Drosophila protein-protein interaction network from FlyPredictome — 1.5 million AlphaFold-Multimer predictions scored with iLIS.
From 1.5 million pairwise predictions, 147,410 heterodimeric PPIs were classified as positive (iLIS ≥ 0.223). To construct this network, we selected positive PPIs supported by fly literature (FlyBase, BioGRID, MIST) or conserved interologs (DIOPT ≥ 4). After filtering for degree ≥ 2 and extracting the giant connected component:
5,224 proteins | 21,657 PPIs | 31 clusters
Leiden community detection identified 31 clusters, each corresponding to a coherent functional theme. Recursive sub-clustering resolved a three-level hierarchy from pathway-level modules down to individual protein complexes.
- Cluster view — global network colored by cluster, zoom and pan
- Grid view — sub-cluster tiles with sub-group coloring
- Evidence filtering — fly only, interolog only, both
- FPR filtering — ≤10%, ≤5%, ≤1%
- Gene search — find any protein and its interactions
- GO enrichment — top terms for BP, MF, and CC per cluster
- Export — download filtered edge lists as CSV
- FlyPredictome Database — predicted structures, PAE maps, iLIS scores
- LIVIA — interactive 3D structure visualization
- AFM-LIS — iLIS analysis pipeline
If you use FlyPredictome in your research, please cite:
Kim, A.-R., Comjean, A., Veal, A., Rodiger, J., Han, M., Hu, Y., & Perrimon, N. (2026). FlyPredictome: A structural atlas of predicted protein-protein interactions in Drosophila. bioRxiv. https://doi.org/10.64898/2026.04.14.718529
@article{kim2026flypredictome,
title = {FlyPredictome: A structural atlas of predicted protein-protein interactions in {Drosophila}},
author = {Kim, Ah-Ram and Comjean, Aram and Veal, Austin and Rodiger, Jonathan and Han, Myeonghoon and Hu, Yanhui and Perrimon, Norbert},
journal = {bioRxiv},
year = {2026},
month = apr,
doi = {10.64898/2026.04.14.718529},
url = {https://doi.org/10.64898/2026.04.14.718529}
}