Sequence similarity networks for the visualisation and exploration of MGnify Proteins

My proposal is to develop a scalable computational pipeline to construct, annotate, and visualise Sequence Similarity Networks (SSNs) for a representative subset of approximately 10 million MGnify proteins. The pipeline will take FASTA files as input and utilise MMseqs2 to efficiently compute all-against-all pairwise sequence similarities. It will then use the Python library NetworkX to construct the mathematical graphs, integrating biological metadata, such as the biome-of-origin, directly into the node attributes to enable contextual environmental exploration. To ensure scalability and avoid memory bottlenecks, the pipeline will leverage NetworkX's backend dispatching architecture to route computationally heavy operations to high-performance GPU libraries like cuGraph. Finally, the graph data will be exported into simple tabular formats compatible with large-scale visualisation platforms like Cytoscape and Cosmograph. The deliverables of this project will include the core Python/Bash SSN generation pipeline, a MGnify metadata annotator, a graph exporter utility, a fully rendered 10-million node reference dataset, and comprehensive documentation alongside an integration test suite.