BSc in Biology, Aristotle University of Thessaloniki. MRes in Bioinformatics, Leeds University, UK. PhD in Bioinformatics, University of Manchester, UK. Post-doctoral researcher, University of Lausanne, Switzerland. EMBO long-term postdoctoral fellow, VIB/UGent, Belgium. Postdoctoral researcher, University of Cambridge, UK. Since 2010, Assistant Professor of Bioinformatics in Genomics and since 2015 as director of the Bioinformatics Research Laboratory, at the Department of Biochemistry and Biotechnology, University of Thessaly, Greece. My research is purely computational and has focused on the evolution of post-translational regulation, like dimerization of transcription factors (bZIP, bHLH and Nuclear Receptor families) and especially post-translational modifications, like protein phosphorylation and methylation. Other major research interests include viral evolution/recombination and primary/secondary metabolism in microbes.
The subject of his speech...
Prediction of post-translational molecular switches and rheostats in proteins
Recent advances in high-throughput (HTP) proteomics have revealed the proteome-wide incidence and emergent crucial role of protein methylation in many cellular processes as well as in cancer and other diseases. Past low-throughput studies focusing on the histone code and especially recent proteome-wide studies have established a strong link between neighboring protein methylation and phosphorylation sites that form molecular switches and/or rheostats/clusters with a digital or analog functional effect. Although a plethora of computational tools exist for the prediction of either protein methylation or phosphorylation sites, they are all trained on datasets of limited size and often questionable quality. In addition, no available tool combines these two predictions in order to identify molecular switches and/or rheostats/clusters at the protein or proteome level. We have mined over 1000 papers to retrieve and stringently filter more than 200 publicly available phosphoproteomic and methylproteomic datasets from diverse eukaryotic species (human, mouse, yeast, Arabidopsis and Toxoplasma gondii) in order to train two neural networks with Keras/Tensorflow. These two networks have been integrated in a webserver that accurately predicts i) methylation and/or phosphorylation sites in eukaryotes ii) the phospho-methyl switches iii) the clusters/rheostats they form and iv) further display them graphically on the protein. Meth-Phos Prometheus can analyze the entire yeast proteome in ~55 minutes.