Objective: To discover microRNA (miRNA)-RNA transcript interactions dysregulated in brains from persons with HIV-associated neurocognitive disorder (HAND), we investigated RNA expression using machine learning tools. Design: Brain-derived host RNA transcript and miRNA expression was examined from persons with or without HAND using bioinformatics platforms. Methods: By combining next generation sequencing, droplet digital (dd)PCR quantitation of HIV-1 genomes, with bioinformatics and statistical tools, we investigated differential RNA expression in frontal cortex from persons without HIV (HIV[-]), with HIV without brain disease (HIV[+]), with HIV-associated neurocognitive disorder (HAND), or HAND with encephalitis (HIVE). Results: Expression levels for 147 transcripts and 43 miRNAs showed a minimum 4-fold difference between clinical groups with a predominance of antiviral (Type I interferon) signaling-, neural cell maintenance-, and neurodevelopmental disorder-related genes that was validated by gene ontology and molecular pathway inferences. Scale of signal-to-noise ratio (SSNR) and biweight midcorrelation (bicor) analyses identified 14 miRNAs and 45 RNA transcripts, which were highly correlated and differentially expressed (p ≤ 0.05). Machine learning applications compared regression models predicated on HIV-1 DNA, or RNA viral quantities that disclosed miR-4683 and miR-154-5p were dominant variables associated with differential expression of host RNAs. These miRNAs were also associated with antiviral-, cell maintenance-, and neurodevelopmental disorder-related genes. Conclusions: Antiviral as well as neurodevelopmental disorder-related pathways in brain were associated with HAND, based on correlated RNA transcripts and miRNAs. Integrated molecular methods with machine learning offer insights into disease mechanisms, underpinning brain-related biotypes among persons with HIV that could direct clinical care.
Read full abstract