Augmenting microarray data with literature-based knowledge to enhance gene regulatory network inference.

Area of Science:

• Systems Biology
• Computational Biology
• Bioinformatics

Background:

• Gene regulatory networks are vital for understanding cellular mechanisms.
• Current computational models often use random gene selection, limiting accuracy.
• Prior knowledge integration in gene network inference is typically restricted.

Purpose of the Study:

• To investigate the impact of using semantic relations from literature to augment microarray data for gene network inference.
• To develop a more accurate model of cellular behavior by incorporating literature-based gene interactions.
• To eliminate the need for random gene selection in non-exhaustive network inference approaches.

Main Methods:

• Augmenting microarray data with semantic relations extracted from scientific literature.
• Employing a genetic algorithm to optimize interaction strengths.
• Utilizing an artificial neural network for fitness evaluation.
• Testing the model on invasive ductile carcinoma of the breast and yeast datasets.

Main Results:

• The developed model demonstrated significantly better fitness compared to state-of-the-art methods relying on random gene selection.
• The inferred gene regulatory networks contained both known and novel relationships, with significant validation of p53 pathway interactions.
• The methodology showed superior performance on both human cancer and yeast datasets.

Conclusions:

• Combining literature-derived semantic relations with microarray analysis effectively generates contribution-weighted gene regulatory networks.
• This integrated approach offers a significant advancement in understanding complex cellular interactions and molecular physiology.
• The method provides a more accurate and biologically relevant representation of gene regulatory processes.