GN/MCL/Kmeans Clustering

Purpose

Based on the STRING database, automatically construct protein interaction networks for gene sets, and output 3 types of clustering results simultaneously.

Running Method

Use the following methods to customize database or use default species database, both can run

SDAS PPI --input gene_300.txt --species human --score_threshold 600 --centers 9 --output results_300
SDAS PPI -i gene_300.txt -o ./result --species human --cluster GN kmeans
SDAS PPI -i gene_300.txt -o ./result --links_db 9606.protein.links.v12.0.txt --aliases_db 9606.protein.aliases.v12.0.txt --cluster GN kmeans

Input Parameter Description

Parameter	Required	Default	Description
-i/--input	Yes		Input gene name list file (gene symbol, one per line)
-o/--output	Yes		Output folder, will be created automatically if not exists
--species	No	human	Species (human/mouse), can be ignored when using custom database，use--links_db and--aliases_db
--links_db	No		Custom protein interaction file path
--aliases_db	No		Custom protein alias file path
--score_threshold	No	700	Protein interaction score threshold, choose between 400-900, higher score means higher reliability, fewer network nodes, default 700
--cluster	No	GN	Clustering algorithm (GN/kmeans/mcl), multiple choices available, default GN
--centers	No	5	kmeans clustering center number, default 5
--inflation	No	2.0	MCL clustering inflation parameter, 1.5-3.0, default 2.0

Output Results Display

Result File	Description
PPI_results.csv	Interaction score between two genes, supports Cytoscape import
cluster_results.csv	Input gene node connectivity and clustering assignment
network_<cluster>_visualization.png/pdf	All gene interaction network diagram, line thickness represents interaction score size, node size represents connectivity, color represents clustering
network_<cluster>_top_clusters.png/pdf	Network diagram of the 9 clusters with the most nodes, circle mode arrangement when node count <50, global arrangement when >50 (separate diagrams for each selected clustering method)

• Protein interaction relationship table: PPI_results.csv Each row represents the interaction score of a pair of genes, can be directly imported into Cytoscape.

  from_gene	to_gene	combined_score
  MEPIA1	CDH17	466
  LGLA3	CDH17	561
  PTK2	CDH17	482

• Clustering result table: cluster_results.csv Each row represents one gene, including its connectivity (how many genes it interacts with), cluster assignment for each clustering algorithm.

  gene	degree	mcl_cluster	kmeans_cluster	betweenness_cluster
  MEPIA1	6	1	2	5
  LGLA3	5	1	2	5

• Interaction network visualization diagram: network_<cluster>_visualization.png/pdf: Shows all gene interaction networks, line thickness between nodes represents interaction score size, same color represents one cluster, node size represents connectivity size. (Separate diagrams for each selected clustering method).

• Maximum clustering subnet circle diagram: network_<cluster>_top_clusters.png/pdf Shows network diagrams of the 9 clusters with the most nodes, circle mode arrangement when node count <50, global arrangement when >50.

Performance Description

Takes a few minutes to run, memory consumption within 1G