Random forest selected transcripts from multiple neurodegenerative diseases reveal common affected cellular processes

Abstract Background Neurodegenerative diseases such as Alzheimer’s disease are associated with changes in specific molecular pathways involving metabolomic and biological processes active in both the central nervous system and vascular peripheral system. Each disease results in neurodegeneration in vulnerable regions and cell types specific to that disease and are expressed in whole blood as disease progresses. Machine learning can be applied to the entire blood transcriptome of neurodegenerative diseases providing information for a pathophysiological relationship and gives us the advantage of using the vast knowledge of gene expression to analyze pathways affected thereby guiding insight into possible treatment strategies. We applied our unique machine learning algorithm to public data sets for selection of transcripts that guide our discovery of affected pathways. Methods We used large public blood microarray mRNA expression datasets, with platforms of more than 30,000 probes covering 10,000 genes, to discover disease transcripts using a novel Random Forest algorithm which selected supervised predictors for six different neurodegenerative diseases including Alzheimer’s disease (AD), Parkinson’s disease (PD), behavioral variant frontotemporal dementia (FTD), amyotrophic lateral sclerosis (ALS), Huntington’s disease (HD), and Friedreich’s ataxia (FRDA). Results Our machine learning selection of transcript disease classifiers reveals that neurodegenerative diseases have common themes as well as unique individual differences. Molecular components which overlap across diseases might be considered neurodegeneration generalities or a result of mixed pathology, while differential components identified by selected feature transcripts may represent unique dysfunctions. Conclusion Our Random Forest algorithm selected important features hidden in mRNA expression values from blood of affected individuals of six neurodegenerative diseases compared to healthy controls. For each of the six diseases studied here, the selected features were categorized by pathway analysis into eight main functional groups. The dysfunction of any or multiple functional groups combined can lead to synapse and cell loss in these neurodegenerative diseases each with specific vulnerabilities in particular tissue and cell types. Vulnerability is usually associated with those cells vulnerable to the disease pathology and are often first to exhibit cell death from cytotoxic events such as neuroinflammation, mitochondrial dysfunction, transcription alterations, apoptosis, protein synthesis dysfunction, cytoskeletal changes, and ubiquitylation/proteasome stalls.