lncAPNet

lncAPNet is an extension of the Activity PASNet (APNet) framework that incorporates long non-coding RNAs (lncRNAs) into explainable, network-based patient stratification and supervised clustering using transcriptomic data.

This repository hosts the implementation and resources for the lncAPNet computational pipeline.

---

Overview

lncAPNet is a novel computational pipeline built upon the recently reported Activity PASNet (APNet) framework
(Gavriilidis et al., 2025).
It performs explainable AI (XAI)-driven supervised clustering of patient cohorts by integrating:

• Transcriptomic analysis
• Biological priors
• The concept of differential activity

The original APNet framework integrates:

• SJARACNe for gene regulatory network inference
• NetBID2 for driver activity analysis
• PASNet, a sparse and interpretable deep learning architecture for patient classification

These components are supported by biological priors and post hoc visualization modules for clinical bioinformatics applications.

---

Extensions Introduced by lncAPNet

lncAPNet expands the APNet toolbox by explicitly modeling lncRNA-driven regulatory mechanisms. The key extensions include:

• Graph-based, nonlinear interpretation of lncRNA–mRNA regulatory relationships using the SJARACNe / NetBID2 framework
• Integration of an lncRNA-specific knowledge graph (lncRNAlyzr-KG) into the biological priors (Evangelista et al., 2025)
• Identification and interpretation of lncRNA-associated regulons derived from NetBID2 analysis
• Enhanced explainability of lncRNA contributions to patient stratification and disease mechanisms

---

Pipeline Status

• Nextflow Installation Instructions

• A Nextflow-based implementation of the lncAPNet pipeline is now available for bulk RNAseq to improve scalability, portability, and reproducibility.

🚧 Post-hoc analyses Python Colab notebooks (connected with Nextflow pipeline) is under active development

🚧 For scRNAseq data is under active development

---

Implementation

Define the local path for input in the Nextflow/pipeline/conf/local.config file. If using singularity, in the same file, uncomment and update the local path to where the singularity containers are stored.

The pipeline should then run using docker or singularity containers with this command from within the Nextflow/pipeline folder:

nextflow run main.nf -c ../conf/local.config -profile singularity,local --outdir output

or, directly passing the input to the CLI command:

nextflow run main.nf -profile docker --input /path/to/data/eset.rds --gene_info /path/to/data/gene_info.xlsx --group0 "GROUP0" --group1 "GROUP1" --comparison "COMPARISON" --outdir output

Example from Nextflow_demo/data toy dataset

nextflow run main.nf -profile docker --input ../Nextflow_demo/data/test_eset.rds --outdir output --gene_info ../Nextflow_demo/data/gene_info.xlsx --group0 "M" --group1 "U" --comparison "IGHV"

bioRxiv link as pre-print:

(Vasileiou V. et al., 2026)