RunDiffusion

Full name SMRUCC.genomics.Analysis.HTS.FELLA.Core.RunDiffusion Assembly SMRUCC.genomics.Analysis.HTS.FELLA Members 4

Network diffusion analysis for FELLA.

The diffusion method uses the regularized Laplacian kernel to propagate perturbation signals from input compounds through the KEGG network.

Mathematical formulation: L_gamma = L + gamma B (regularized Laplacian) K = L_gamma^{-1} (diffusion kernel) f = K y (diffusion scores)

where: L = D - A (unnormalized graph Laplacian) D = diagonal degree matrix A = adjacency matrix B = diagonal "leak" matrix (1 for pathway nodes, 0 otherwise) gamma = regularization parameter (default 1.0) y = input vector (1 for input compounds, 0 otherwise)

P-score computation (normality approximation): z_i = (f_i - mu_i) / sigma_i p_i = Phi(z_i)

where mu_i and sigma_i are the mean and standard deviation of the null distribution for node i, computed analytically.

01 Syntax

SMRUCC.genomics.Analysis.HTS.FELLA.Core.RunDiffusion

02 Methods

Name	Overloads	Summary
Run	1	Run diffusion analysis with specified approximation method.
ComputePScoresNormality	1	Compute p-scores using the normality approximation.
ComputePScoresSimulation	1	Compute p-scores using Monte Carlo simulation.
user_InputCount	1	Count non-zero entries in input vector

03 Members

method Run #

Run(FellaUser, FellaData, ApproximationMethod, Int32)

Run diffusion analysis with specified approximation method.

Parameters

Name	Type	Description
`user`	FellaUser	FELLA user object with defined input compounds
`data`	FellaData	FELLA data object with precomputed diffusion kernel
`approx`	ApproximationMethod	Approximation method for p-value computation
`nSim`	`Int32`	Number of Monte Carlo iterations (only used if approx = Simulation)

Returns

Enrichment result with diffusion scores and p-scores for all nodes

method ComputePScoresNormality #

ComputePScoresNormality(Double(), FellaData, EnrichmentResult)

Compute p-scores using the normality approximation. For each node i: z_i = (f_i - mu_i) / sigma_i p_i = Phi(z_i) where mu_i and sigma_i are precomputed null distribution statistics.

method ComputePScoresSimulation #

ComputePScoresSimulation(Double(), Double(), FellaData, EnrichmentResult, Int32)

Compute p-scores using Monte Carlo simulation. For each permutation:

Randomly sample nInput compounds from the background
Compute diffusion scores for the random input
Record the scores

P-score for node i = (count of null scores >= observed + 1) / (nSim + 1)

method user_InputCount #

user_InputCount(Double())

Count non-zero entries in input vector