Learning from a genetic cornucopia Over the last years, computation-intensive methods are increasingly used to analyse biomedical signals. The general approach falls under the rubric of artificial intelligence in which computer programs mimic the human brain's organisation to 'learn' important features from the data set. Industrial Technologies © Theodore Alexandrov To date, computers can process information serially and solve complex problems far faster than humans believed possible. Expert systems have surpassed human skills for a broad range of tasks. However, qualitative functions that the brain performs naturally by recognising patterns to make judgements are still extremely difficult for computers. Within the EU-funded project ARSINFORMATICA (Artificial intelligence, branching processes and coalescent – Searching the information from a genetic cornucopia), scientists explored the capabilities of machine learning to comb through genomic data. The aim was to find links between genotypes and diseases and understand natural selection. Mathematical models were developed to study how gene barriers could have remained between anatomically modern humans and Neanderthals. The period of coexistence of Neanderthals and Homo Sapiens was the basis of this intriguing problem. The scientists looked into the interbreeding between the two populations to confirm the findings of recent DNA-based studies. In addition, signatures of natural selection at the molecular level were searched in single nucleotide polymorphism data. The results obtained from widely used approaches are obscured by the presence of hypotheses, like population growth and geographic substructure. The scientists dealt with this problem by the application of multiple null hypotheses that assume different population scenarios. The capabilities of the so-called multi-null-hypotheses method as knowledge generator were demonstrated in the search for natural selection in genes implicated in human familial cancers. Although it is computationally demanding and therefore cannot be applied for a similar search in many genes, it proved able to evaluate the influence of time changes and demography in population size. Scientific papers with the results have been published in high-impact peer-reviewed journals and presented at international conferences. The ultimate objective of the ARSINFORMATICA project was to reinforce the international dimension of research on computational biology and bioinformatics in Poland. Collaborative links were also established with the United States. Facilitating long-term collaborative research at the international level will contribute to strengthening the European Research Area (ERA). Machine learning techniques developed and translated into practical tools are expected to find applications beyond human genetics. In particular, machine learning algorithms together with data mining processes can help in pattern recognition even with unstructured data. Keywords Genetic cornucopia, artificial intelligence, ARSINFORMATICA, machine learning, genomic data, natural selection
