Machine learning identification of molecular targets for medulloblastoma subgroups using microarray gene fingerprint analysis

The study introduces a structured methodology for the identification of molecular targets that accurately classify medulloblastoma subgroups: WNT, SHH, Group 3 (G3) and Group 4 (G4). An artificial neural network (ANN) model trained on microarray gene expression data determined minimal gene combinations for each subgroup. The classification achieved an average accuracy of 96%, demonstrating the effectiveness of the proposed approach. Feature selection using the Kruskal-Wallis and <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:msup><mml:mrow><mml:mi>χ</mml:mi></mml:mrow> <mml:mrow><mml:mn>2</mml:mn></mml:mrow> </mml:msup> </mml:math> tests revealed statistically relevant genes contributing to subgroup discrimination. Reverse transcription followed by digital Polymerase Chain Reaction (dPCR) measured the expression levels of a subset of these genes in tumor samples, validating the computational predictions with experimental evidence. The integration of machine learning and molecular quantification provides a reproducible framework for medulloblastoma subgroup classification supported by both statistical and experimental consistency.