Developmental gene regulatory network connections predicted by machine learning from gene expression data alone.

Area of Science:

• Developmental Biology
• Computational Biology
• Systems Biology

Background:

• Gene regulatory network (GRN) inference is crucial for understanding developmental processes.
• Embryonic development presents unique challenges for GRN prediction due to its dynamic and complex nature.
• Machine learning offers new avenues to complement traditional experimental methods in GRN construction.

Purpose of the Study:

• To demonstrate the efficacy of the Priors Enriched Absent Knowledge (PEAK) algorithm for inferring gene regulatory networks from gene expression data alone.
• To assess the performance of PEAK in predicting developmental GRNs using time-series expression data.
• To generate novel predictions for gene interactions within the sea urchin developmental GRN.

Main Methods:

• Utilized the Priors Enriched Absent Knowledge (PEAK) network inference algorithm, a noise-robust method employing ordinary differential equations and Elastic Net.
• Applied PEAK to two gene expression datasets from the purple sea urchin, Stronglyocentrotus purpuratus.
• Validated predictions against established GRN models derived from extensive experimental data.

Main Results:

• Achieved high sensitivity (up to 81.58%) in identifying known gene interactions within the sea urchin developmental GRN.
• Generated novel, experimentally supported predictions for previously undescribed gene interactions.
• Demonstrated that GRN predictions matching known interactions are achievable using only time-series gene expression data.

Conclusions:

• The PEAK algorithm successfully infers gene regulatory networks from gene expression data during embryonic development.
• Gene expression time-series data alone is sufficient for accurate GRN prediction in developmental contexts.
• The study provides a valuable resource of novel GRN predictions for future research in sea urchin development.