Gene set creation algorithm for microarray studies with low sample size

This study aims to validate a novel approach for discovering cell-specific gene sets based on gene expression profiles. More than a billion people suffer from neurological disorders, and understanding how cells interact in the context of different pathologies could lead to viable medical interventions. To this end, the field requires reliable biological markers to discriminate distinct cell types present in a heterogeneous population. Distilling these markers is complex, hampered by problems such as the curse of dimensionality when relating large counts of genes to few cell type observations. This research presents a method for overcoming these challenges in a dataset of microarray gene expression values. A bootstrapped LDA algorithm was used to construct new gene sets and validated using GSVA. Gene sets from a non-bootstrapped LDA process provided a baseline to compare gene set performance in the face of genetic noise. Results showed a marked improvement in half of the validation examples, while the other sets were not significantly enriched.