lncAPNet

lncAPNet integrates graph-based nonlinear inference of lncRNA-mRNA interactions using NetBID2's and scMINER's activity logic within a lncRNA-focused SJARACNe co-expression network, coupled with PASNet, a biologically informed sparse deep learning model, to enable explainable identification of lncRNA drivers in patient transcriptomic data.

Key Features:

• Graph-based nonlinear inference: Infers lncRNA-mRNA interactions using NetBID2's and scMINER's activity logic, moving beyond descriptive linear co-expression analyses.
• lncRNA-focused co-expression network: Constructs networks using SJARACNe specifically focused on long non-coding RNAs.
• Biologically informed sparse deep learning: Employs PASNet, a sparse deep learning model incorporating biological priors to address data sparsity, noise, and poor interpretability limitations of emerging ML/DL methods.
• Explainable driver identification: Enables interpretable discovery of lncRNA drivers and lncRNA-mediated programs in cancer progression.
• Multi-omics compatibility: Supports both bulk RNA-seq and single-cell RNA-seq datasets, as demonstrated by integration of bulk RNA-seq and scRNA-seq data.

Scientific Applications:

• Cancer transcriptomics: Identifies lncRNA drivers in cancer progression through analysis of patient transcriptomic data, as demonstrated in Chronic Lymphocytic Leukemia, Prostate Adenocarcinoma, and Breast Invasive Carcinoma case studies.
• lncRNA-mRNA regulatory motif discovery: Uncovers systems-level lncRNA-mRNA interaction motifs that illuminate lncRNA-mediated programs in disease.
• Multi-platform RNA-seq analysis: Applicable to both bulk RNA-seq datasets and integrated bulk RNA-seq with scRNA-seq omics datasets.

Methodology:

lncAPNet extends the APNet workflow by integrating graph-based nonlinear inference of lncRNA-mRNA interactions using NetBID2's and scMINER's activity logic within a lncRNA-focused SJARACNe co-expression network, coupled with PASNet, a biologically informed sparse deep learning model.