Candidate genes associated with neurological manifestations of COVID-19: Meta-analysis using multiple computational approaches

COVID-19 develops certain neurological symptoms, the molecular pathophysiology of which is obscure. In the present study, two networks were constructed and their hub-bottleneck and driver nodes were evaluated to consider them as ‘target genes’ followed by identifying ‘candidate genes’ and their associations with neurological phenotypes of COVID-19. A tripartite network was first constructed using literature-based neurological symptoms of COVID-19 as input. The target genes evaluated therefrom were then used as query genes to identify the co-expressed genes from the RNA-sequence data of the frontal cortex of COVID-19 patients using pair-wise mutual information to genes. A ‘combined gene network’ (CGN) was constructed with 189 genes selected from TN and 225 genes co-expressed in COVID-19. Total 44 ‘target genes’ evaluated from both networks and their connecting genes in respective networks were analyzed functionally by measuring pair-wise ‘semantic similarity scores’ (SSS) and finding Enrichr annotation terms against a set of genes. A new integrated ‘weighted harmonic mean score’ was formulated using SSS and STRING-based ‘combined score’ to select 21 gene-pairs among ‘target genes’ that provided 21 ‘candidate genes’ with their properties as ‘indispensable driver nodes’ of CGN. Finally, six pairs providing seven prevalent candidate genes (ADAM10, ADAM17, AKT1, CTNNB1, ESR1, PIK3CA, FGFR1) exhibited direct linkage with the neurological phenotypes under tumour/cancer, cellular signalling, neurodegeneration and neurodevelopmental diseases. The other phenotypes under behaviour/cognitive and motor dysfunctions showed indirect associations with the former genes through other candidate genes. The pathophysiology of ‘prevalent candidate genes’ has been discussed for better interpretation of neurological manifestation in COVID-19.