MORPH

MORPH identifies candidate pathway genes by integrating known gene sets, expression profiles, interaction and metabolic networks with machine learning to rank genes associated with specific biological pathways.

Key Features:

• Evidence integration: Integrates known gene sets from target pathways, expression profiles, interaction networks, metabolic networks, and high-throughput experimental data to inform predictions.
• Machine learning selection: Employs machine learning to evaluate various combinations of datasets and select the optimal combination for each analysis.
• Ranked candidate output: Produces a ranked list of candidate genes predicted to be part of the specified pathway.
• Input specialization: Accepts input comprising known genes from the target pathway and relevant biological networks.
• Cross-validation evaluation: Demonstrated high-quality cross-validation performance on pathway datasets.
• Experimental corroboration: High-ranked Arabidopsis genes for photosynthesis light reactions, homogalacturonan biosynthesis, and chlorophyll biosynthetic pathways have experimental verification.
• Cross-pathway detection: Identifies genes from related pathways that share precursors or exhibit coregulation, as shown for the carotenoid pathway in Arabidopsis and tomato (Solanum lycopersicum).
• Species- and pathway-scale testing: Validated on 230 known pathways in Arabidopsis thaliana and 93 known pathways in tomato (Solanum lycopersicum).

Scientific Applications:

• Novel gene discovery: Prioritizes candidate genes for discovery of previously unknown members of biological pathways.
• Experimental prioritization: Ranks genes to guide experimental validation efforts in pathway biology.
• Cross-species pathway analysis: Applies to comparative pathway studies in Arabidopsis thaliana and Solanum lycopersicum.
• Pathway crosstalk and coregulation analysis: Detects genes involved in related pathways that share metabolic precursors or regulatory coexpression.
• Method benchmarking: Enables assessment of pathway membership prediction through cross-validation on curated pathway sets.

Methodology:

Accepts known pathway genes and biological networks as input, integrates known gene sets, expression profiles, interaction and metabolic networks (and high-throughput experimental data), applies machine learning to evaluate and select optimal dataset combinations, and outputs a ranked list of candidate pathway genes, with performance assessed by cross-validation.