Understanding Complex Diseases and Disease Causative Agents : The Machine Learning way

Abstract: Diseases can be caused by foreign agents – pathogens – such as viruses, bacteria and other parasites, entering the body or by an internal malfunction of the body itself. The partial understanding of diseases like cancer and the ones caused by viruses, like the influenza A viruses (IAVs) and the human immunodeficiency virus, means we still do not have an efficient cure or defence against them. In this thesis we developed and applied combinations of machine learning strategies to study some of the deadliest human diseases and the agents causing them. The results obtained in this study further our understanding about them, paving the way for the development of more efficient and more reliable counter strategies against them.In Paper I we studied the genetic make up of the highly pathogenic (HP) avian influenza viruses and identified a viral genetic background that could potentially transform a low pathogenic (LP) strain into HP. In Paper II we identified combinatorial signatures in the IAVs genome that potentially could affect their adaptation to humans.Candidate HIV vaccine studies are usually carried out in nonhuman primate models. In Paper III we analysed the host responses of immunized Rhesus Macaques against the simian immunodeficiency virus infection. We found that protection in Rhesus Macaques is mediated by a gradually built up immune response, in contrast to a strong initial immune response, which we found to be detrimental to protection.In Paper IV we analysed 9 different cancer types and identified 38 novel long noncoding RNAs (lncRNAs) that have a disrupted expression in multiple cancer types – pan-cancer differentially expressed (DE) lncRNAs. In addition, we also found 308 novel lncRNAs whose dysregulation was specific to a certain cancer type (cancer-specific DE lncRNAs).