FlashPPI in SeqHub: Proteome-Scale Protein-Protein Interaction Prediction

Protein-protein interactions (PPIs) govern nearly every process in microbial biology: how cells regulate metabolism, how pathogens establish infection, how proteins assemble into functional complexes. Yet for most microbial genomes, the vast majority of these interactions remain uncharacterized. That gap isn't just an inconvenience; it's a fundamental limit on what researchers can discover. When you don't know a protein's partners, it's hard to know what it does, how it fits into a pathway, or what happens when it's disrupted. Gaps in protein interaction networks also directly limit protein function prediction, leaving hypothetical and unannotated proteins functionally invisible.

The reason so many interactions remain unknown comes down to a simple scaling problem: predicting PPIs across a full proteome has historically required months of compute time, forcing researchers to investigate only a handful of interactions at a time rather than screening genome-wide protein interaction networks.

FlashPPI is a new open-source model from Tatta Bio (read the paper) that addresses the proteome-scale PPI bottleneck in two key ways: it screens an entire microbial proteome at once rather than comparing proteins pairwise, and it does so in minutes rather than months by working directly from protein sequences rather than predicted structures. That combination of full-proteome coverage at linear-time speed is what makes proteome-wide PPI screening practical as a routine part of a research workflow for the first time.

Today we're introducing FlashPPI in SeqHub — proteome-wide PPI prediction, directly in your browser.

FlashPPI delivers a four-fold performance improvement over existing sequence-based methods, with interaction interface predictions that match the quality of state-of-the-art structure folding models. For the first time, a complete map of predicted PPIs across your microbial genome is a standard part of your SeqHub workflow, not a months-long side project. And because FlashPPI is built into SeqHub, interaction results sit alongside functional annotation, genomic context visualization, co-occurrence search, and structure prediction, giving you a richer, more connected picture of your genome from a single run.

Accelerate functional annotation

FlashPPI results in SeqHub are integrated with functional annotation, genomic context visualization, co-occurrence search, and structure prediction, giving you a richer, more connected picture of your genome from a single run.

Shed light on uncharacterized proteins

Many microbial genomes, especially from environmental and metagenomic samples, are full of hypothetical or functionally unannotated proteins. Knowing which other proteins they interact with is often the first clue to what they do. With FlashPPI, you can get interaction predictions for every protein in your genome at once, rather than investigating candidates one at a time.

Map protein complexes across your proteome

Multi-protein complexes drive virtually every biological process. Proteome-wide interaction screening gives you a systems-level view of how proteins assemble into functional units — a level of analysis previously only tractable for well-studied model organisms.

Investigate microbial community and host-pathogen dynamics

For researchers studying how microbes interact with each other or with a host, a proteome-wide PPI map adds a new layer of molecular context for understanding those relationships.

Running a proteome-wide PPI screen in SeqHub takes three steps:

1. Upload your genome to SeqHub, or select one you've already analyzed
2. Click the PPI action in the navigation bar to run FlashPPI across the full proteome
3. Explore the full interaction network, click into specific clusters to dive deeper, and view contact maps showing how two proteins interact at the residue level, with interacting regions highlighted in blue

No local infrastructure, no setup, no separate compute resources required.

Not ready to run your own genome yet? We've put together a preview dataset of FlashPPI outputs inside SeqHub so you can explore what the results look like, including the interaction network, interface predictions, and functional annotations, before running your own analysis.

Explore the FlashPPI preview dataset →

FlashPPI is described in full in our preprint, now available on bioRxiv. The model is also fully open source — find the code, training details, and documentation on GitHub.