Transcriptome Derived Artificial neural networks predict PRRC2A as a potent biomarker for epilepsy.

Epilepsy refers to the occurrence of two or more than two reiterative seizures. The occurrence of seizure is governed by the excessive electrical discharges in the cortex of the brain. Bioinformatics is crucial in diagnosing, prognosticating, and treating neurological disorders. It uses methodologies, computational tools, software, and databases to probe disease molecular underpinnings and identify biomarkers. It aids clinicians in addressing patient parameters and translational research. Artificial neural networks (ANNs) are computer models that attempt to mimic the neurons present in the human brain. This computerized neuronal model is used for analyzing and comprehending large and complex data sets. In the present study, three GEO datasets (GSE190451, GSE140393, and GSE134697) were retrieved from NCBI for the identification of differentially expressed genes using the DESeq2 package. The study identified 7 up-regulated genes (PRRC2A, FCGR3B, HLA-DRB, ENSG00000280614, ENSG00000281181, SLN, C4A) in patients with epilepsy. Furthermore, WEKA software was used for feature selection and classification of DEGs using feature selection algorithms namely Correlation Feature Selection, ReliefF, and Information Gain and classification methods such as Logistic regression, Classification via regression, Random forest, Random subspace, and Logistic model trees. After the analysis, out of the 7 genes, the C4A gene was removed as it yielded the lowest feature selection statistics. Lastly, R Studio was used for constructing the Artificial Neural Network of the 6 identified DEGs. The model's performance was evaluated using the "pROC" R package, and an AUC of 0.720 was obtained, indicating that the model had excellent classification accuracy. The NeuralNet package of R revealed that PRRC2A had the highest generalized weight value indicating the increased expression of these genes when all other parameters are constant. Therefore, PRRC2A can be used as a potential biomarker for the diagnosis of epilepsy.