GAM-clustering analysis

Performs the core GAM-clustering procedure to identify transcriptionally coordinated metabolic modules. Starting from initial expression patterns, the algorithm iteratively scores genes, solves SGMWCS optimization problems on the metabolic network, and updates module-specific expression profiles until convergence. Additional refinement steps merge highly similar modules, remove uninformative patterns, and control module size. The function returns final metabolic modules, scored networks, module expression patterns, and detailed iteration statistics.

Usage

gamClustering(
  E.prep,
  network.prep,
  cur.centers,
  start.base = 0.5,
  base.dec = 0.1,
  max.module.size = 50,
  cor.threshold = 0.8,
  p.adj.val.threshold = 0.001,
  batch.solver = seq_batch_solver(solver),
  work.dir,
  show.intermediate.clustering = TRUE,
  verbose = TRUE,
  collect.stats = TRUE,
  reference.patterns = NULL
)

Arguments

E.prep

Expression matrix after the prepareData() function.

network.prep

Network edge table driven from prepareNetwork() function.

cur.centers

Initial patterns produced by preClustering() function.

start.base

The parameter which influences modules sizes.

base.dec

The value controlling how strongly base parameter should be reduced if some module's size is bigger that max.module.size. The update rule is: base <- base * (1 - base.dec). Detaulf: 0.1.

max.module.size

Maximal number of unique genes in the final module.

cor.threshold

Threshold for correlation between module patterns.

p.adj.val.threshold

Padj threshold of geseca score for final modules.

batch.solver

Solver for SGMWCS problem.

work.dir

Working directory where results should be saved.

show.intermediate.clustering

Whether to show or not heatmap of intermediate clusters.

verbose

Verbose running.

collect.stats

Whether to save or not running statistics.

reference.patterns

Matrix of reference patterns to track correlation of centers against. Pattern per row. Number of columns should be the sames as in E.prep.

Value

results$modules – Metabolic modules.

results$nets – Scored networks.

results$patterns.pos – Modules' patterns (genes with positive score only considered).

results$patterns.all – Modules' patterns (all genes considered).

results$iter.stats – Statistics from iterations.

Parameter tuning

There is a set of parameters which determine the size and number of the final modules. We recommend starting with the default settings, but the following adjustments may be useful:

• If you consider final modules to be too small or too big and it complicates interpretation for you, you can either increase or reduce by 10 units the max.module.size parameter.

• If among final modules you consider presence of any modules with too similar patterns, you can reduce by 0.1 units the cor.threshold parameter.

• If among final modules you consider presence of any uninformative modules, you can reduce by 10 times the p.adj.val.threshold parameter.