GMCLDA

GMCLDA predicts potential associations between long non-coding RNAs (lncRNAs) and diseases by applying geometric matrix completion to integrated lncRNA functional, disease phenotypic, and sequence similarity data for association inference.

Key Features:

• Geometric Matrix Completion: Infers missing entries in the lncRNA-disease association matrix by exploiting functional similarities among lncRNAs and phenotypic similarities among diseases.
• Sparsity and Robustness: Imposes sparsity constraints on predicted values to increase robustness of association estimates.
• Disease Semantic Similarity (Disease Ontology): Computes disease semantic similarity using the Disease Ontology (DO) hierarchy to quantify phenotypic relationships between diseases.
• lncRNA Gaussian Interaction Profile Kernel Similarity: Calculates lncRNA similarity from known interaction profiles using a Gaussian interaction profile kernel to capture functional relationships.
• Sequence Similarity Measurement (Needleman-Wunsch & K-nearest neighbors): Measures sequence similarity with the Needleman-Wunsch algorithm and pre-fills interaction profiles for new lncRNAs using K-nearest neighbors defined by this similarity.

Scientific Applications:

• Disease Mechanism Understanding: Supports elucidation of molecular disease mechanisms by predicting lncRNAs potentially involved in disease processes.
• Diagnosis and Treatment Biomarker Prioritization: Prioritizes candidate lncRNAs as potential biomarkers or therapeutic targets to inform diagnostic and treatment research.
• Candidate Prioritization and Case Studies: Enables case-study applications such as inferring potential lncRNAs associated with renal cancer for downstream experimental validation.

Methodology:

Integrates lncRNA functional similarities, disease semantic similarity from Disease Ontology, and sequence information (Needleman-Wunsch for K-nearest neighbors prefill) into an lncRNA-disease association matrix and applies geometric matrix completion with sparsity constraints to estimate missing associations.