Skip to main content

Table 3 Unknown cell-type clustering performance of the different models analyzed for the LPGO experiments (P = 4). Although our pathway-based models are nearly ten times smaller (sparse), the performance is very close to the PPI-based NN. The mean of 20 splits was reported

From: Integrating pathway knowledge with deep neural networks to reduce the dimensionality in single-cell RNA-seq data

Architecture

Number of nodes

(2nd hidden layer)

Homogeneity

Completeness

V-measure

ARI

AMI

Fowlkes-Mallows

Average

Dense

–

0.801

0.799

0.798

0.725

0.786

0.814

0.787

Dense with pathways

–

0.804

0.797

0.798

0.718

0.786

0.811

0.786

Dense with PPI

–

0.811

0.804

0.805

0.728

0.794

0.817

0.793

Dense with PPI and GRN

–

0.820

0.808

0.812

0.746

0.802

0.827

0.802

Signaling pathways

–

0.797

0.788

0.790

0.716

0.778

0.809

0.780

Signaling pathways

100

0.775

0.803

0.786

0.729

0.774

0.820

0.781