Machine Learning and Bioinformatics Models to Identify the Genetic Link of Neurological Diseases Associated With the Causal Risk Factors

Abstract

Neurological diseases (NDs) are causing burgeoning burden to the patients, healthcare sector and the entire society. Hundreds of millions of people are currently a ected by various NDs worldwide and the number is increasing very rapidly. The most frequent categories include Alzheimer's disease (AD), Parkinson's disease (PD), epilepsy, Multiple sclerosis (MS), stroke and other cerebrovascular disorders, migraine and other headache, malignant brain tumors such as Glioblastoma multiforme (GBM) etc. Despite numerous research initiatives, preventive and therapeutic options for most of these NDs still remain very limited. Taking AD as an example, fully e ective preventive strategies are unavailable till now. Preventive measures usually comprise primary prevention based on risk reducing by identifying in uential factors and secondary prevention through early detection and abatement of the disease at initial stage. But the inadequate epidemiological knowledge of AD risk factors and absence of early premortem accurate diagnosis has foiled AD prevention. However, better insight about the co-occurrence of other neurological complications with AD can yield preventive and therapeutic advancement. On the other hand, enhanced understanding about the factors that impacts the response to the treatment could prolong the survival period. For instance, GBM is such an ND with shorten survival period provided the rst line treatment include brain surgery followed by chemotherapy and radiotherapy. In this context, research initiatives to mitigate the information gap regarding how the causative factors a ect the cell pathways altered in NDs and their comorbidities can alleviate the disease burden. Availability of high throughput technologies including microarray and next-generation sequencing (NGS) of tissue mRNA to analyse large-scale transcriptomic data have excelled various bioinformatics methodologies as promising tools in biomedical research eld. These approaches include di erential gene expression analysis, protein-protein interactions (PPIs), gene ontology (GO), metabolic pathway and regulatory factor analysis. Genetic inspection into the transcriptomic data through these tools yields better insight into the molecular pathogenesis of any health condition in junction with its causative factors and related complications. In addition to this, the exponentially increasing amount of accessible biological data has made machine learning techniques as promising means of discovering hidden genetic knowledge. In this thesis, we presented bioinformatics and computational frameworks based on transcriptomic data and machine learning based survival prediction models.