ST-Steiner: a spatio-temporal gene discovery algorithm.

Area of Science:

• Genetics
• Neuroscience
• Computational Biology

Background:

• Autism spectrum disorder (ASD) genetic architecture is complex, with whole exome sequencing (WES) identifying limited risk genes.
• Existing network-based gene discovery methods use static networks, failing to account for the dynamic nature of neurodevelopment.

Purpose of the Study:

• To develop a spatio-temporal gene discovery algorithm that incorporates evolving gene co-expression networks during neurodevelopment.
• To improve the identification of ASD risk genes by modeling the dynamic changes in gene interactions.

Main Methods:

• Developed a novel algorithm based on a prize-collecting Steiner forest problem applied to evolving gene co-expression networks.
• Adapted the algorithm to model neurodevelopment and transfer information across different developmental windows.
• Applied the algorithm to ASD WES data using BrainSpan co-expression networks from early and mid-fetal periods.

Main Results:

• Identified novel risk clusters for ASD by analyzing spatio-temporal gene interactions.
• Demonstrated that incorporating the temporal dimension significantly increases predictive power on an independent dataset.
• Achieved higher enrichment in ASD-related functions for predicted clusters compared to state-of-the-art methods.

Conclusions:

• The proposed spatio-temporal algorithm enhances gene discovery for complex disorders like ASD.
• Accounting for the dynamic nature of neurodevelopment is crucial for accurate gene association studies.
• The algorithm provides interpretable results and improves the identification of functional gene clusters.