Új gyógyszercélpontok azonosítása poligénes betegségekben mesterséges intelligencia algoritmusok segítségével


Intézmény: Semmelweis Egyetem
gyógyszerészeti tudományok
Gyógyszertudományok Doktori Iskola

témavezető: Petschner Péter
helyszín (magyar oldal): Semmelweis Egyetem
helyszín rövidítés: SE DI

A kutatási téma leírása:

Artificial intelligence (AI) has begun to transform our lives. Self-driving cars, chat bots or personal assistants on our phones demonstrate the most visible use of the technology. Less well-known is that AI algorithms can support drug discovery and development. Especially tempting is the use of such algorithms in finding novel drug targets and the better characterization of polygenic disorders, like depression or migraine.
There are several reasons, why AI can provide the largest benefit in the research of such disorders: 1) the small individual and heterogeneous contribution of genetic variants in such polygenic disorders represents a problem in genetic analyses (Flint and Kendler, 2014; Petschner, Gonda et al., 2018) and in addition, it is relatively hard to find significant genetic factors due to the small n big p statistical problem with classic methods (Petschner, Bagdy et al., 2015); 2) besides of the heterogeneous genetic background environmental factors may alter disease etiology and course (Gonda, Petschner et al., 2018); 3) comorbid diseases can interact with underlying disease promoting factors (Gonda, Petschner et al., 2018); and 4) these disease terms are heterogeneous and most probably depression or migraine can be considered umbrella terms debating the homogeneity of the diagnoses (see e.g. (Fried and Nesse, 2015)).
All these problems suggest that substantial improvement in statistical and analytic approaches is required, which consider multiple factors in disease development, including genetic, transcriptomic, phenomic, environmentomic and comorbid ones. While such a complex approach is only wishful thinking today, AI algorithms are able to spot relevant patterns behind such complex problems. Additional advantage is that these polygenic disorders are common in the population and the application of AI requires large datasets, which are hardly obtained for other, less common diseases.
Years ago a collaboration between the lab and the Department of Measurement and Information Systems at the Budapest University of Technology and Economics was established for the development of novel AI-based methods in the analysis of depression’s background. This collaboration is now beginning to bear fruit. A paper investigating the comorbidity map of major depression on the UK Biobank dataset (Marx, Antal et al., 2017) and another one addressing the role of different genes behind depression were recently published (Gonda, Hullam et al., 2018). In both of them, Bayesian statistics-based AI algorithms formed the backbone of the calculations. Nevertheless, we are constantly developing our methods and in a work in progress led by the supervisor that investigates migraine genetics, we use a deep neural network based classifier in addition to the Bayesian one.
Students can participate in these projects and help in the collection of data, calculations or interpretation of results in addition to the development of methodology. Those applying will be provided with the opportunity to learn the basics of AI algorithms, the use of deep neural networks based classifiers and Bayesian statistics based AI methods on our large datasets of multiple thousands subjects (NEWMOOD, UK Biobank) and the investigation with these methods of the genetic, gene-environment, gene-gene and gene-phenotype networks behind migraine and depression. The final aim is to discover novel drug targets, mechanisms and predictive factors in these disorders and their related phenotypes, and also, to identify novel patient subgroups, in whom current therapies may be more effective.

Flint, J., and Kendler, K.S. (2014). The genetics of major depression. Neuron 81, 484-503.
Fried, E.I., and Nesse, R.M. (2015). Depression is not a consistent syndrome: An investigation of unique symptom patterns in the STAR*D study. J Affect Disord 172, 96-102.
Gonda, X., Hullam, G., Antal, P., Eszlari, N., Petschner, P., Hokfelt, T.G., Anderson, I.M., Deakin, J.F.W., Juhasz, G., and Bagdy, G. (2018). Significance of risk polymorphisms for depression depends on stress exposure. Sci Rep 8, 3946.
Gonda, X., Petschner, P., Eszlari, N., Baksa, D., Edes, A., Antal, P., Juhasz, G., and Bagdy, G. (2018). Genetic variants in major depressive disorder: From pathophysiology to therapy. Pharmacol Ther.
Marx, P., Antal, P., Bolgar, B., Bagdy, G., Deakin, B., and Juhasz, G. (2017). Comorbidities in the diseasome are more apparent than real: What Bayesian filtering reveals about the comorbidities of depression. PLoS Comput Biol 13, e1005487.
Petschner, P., Bagdy, G., and Tothfalusi, L. (2015). [The problem of small "n" and big "P" in neuropsycho-pharmacology, or how to keep the rate of false discoveries under control]. Neuropsychopharmacol Hung 17, 23-30.
Petschner, P., Gonda, X., Baksa, D., Eszlari, N., Trivaks, M., Juhasz, G., and Bagdy, G. (2018). Genes Linking Mitochondrial Function, Cognitive Impairment and Depression are Associated with Endophenotypes Serving Precision Medicine. Neuroscience 370, 207-217.

Jelentkezési határidő: 2021-02-20

Minden jog fenntartva © 2007, Országos Doktori Tanács - a doktori adatbázis nyilvántartási száma az adatvédelmi biztosnál: 02003/0001. Program verzió: 2.2358 ( 2017. X. 31. )