Introducing Gnosis Data Analysis IKE Mens Ex Machina Group www.gnosisda.com mensxmachina.org
Ioannis Tsamardinos o Education o B.Sc. from CSD, University of Crete, 1991-1995 o M.Sc., Ph.D. Intelligent Systems Program, University of Pittsburgh, 1995 - 2001 o Internship and Award, NASA Ames center o Previous Positions o Assistant Professor, Department of Biomedical Informatics, Vanderbilt University, 2001- 2006 o Current Positions o Professor, CSD, UoC o Affiliated Faculty, Institute of Applied Mathematics, FORTH o Visiting, Huddersfield University o Founder, CEO, Gnosis Data Analysis 2
Recognition o 110+ papers, 7000+ citations, 600 citation/year o ~25 funded grants o NASA Group Achievement Award, NASA Ames o ERC, ARISTEIA II grantee o Keynote and invited speaker o Best paper awards 3
Mens Ex Machina o Ο από μηχανής νούς o 4 post-docs o 1 Ph.D. candidate o ~10 M.Sc. o Undergrads o 1 administrator! 4
Past MXM students and members o Assistant Professor, University of Pittsburgh o Google, Zurich o Assistant Professor, Rethymnon o Researcher, FORTH 5
Research Topics o Machine Learning , Data Science, and Artificial Intelligence o Feature Selection, Causal Discovery, Automated Machine Learning (AutoML) o Bioinformatics and Biomedical Informatics o Interdisciplinary research : medicine, biology, materials science 6
Research Types Theory and Math Philosophy Algorithms Applications for new Tools and Systems biomedical knowledge
Learning Causal Models o Google “classic” of 2006 in the field of Artificial Intelligence o ERC Consolidator Award CAUSALPATH 8
Big Data Feature Selection o Select the quantities that are most predictive of an outcome, in combination o New algorithm scaling to 40M quantities and infinite sample size 9
Figure 4. Predicted causal pairs for CLCD. CLCD outputs 82 different predictions (38 unique cause-effect pairs. CLCD results for the remaining subpopulations are grouped into CD14+ surface- and monocytes (CD14+HLA − Dr − , CD14+HLA − Dr high , CD14+HLA − Dr mid , CD14 − HLA − Dr − , CD14 − HLA − Dr high , CD14 − HLA − Dr mid ) and B cells (IgM − and IgM+). Edge thickness corresponds to frequency of appearance in different contexts (activators, inhibitors, subpopulations where applicable). Green edges are confirmed in at least one KEGG pathway, while brown edges are found reversed in at least one KEGG pathway.
✓ University of Crete spin-off ✓ Headquarters and R&D in Heraklion ✓ Greek, Danish, German partners ✓ Decades long-experience in applied machine learning in life sciences ✓ 12 employees + 3 employees + new positions
International Partnerships 12
JAD BI 5 Step Analysis Apply to New Data https://www.jadbio.com/ 13
Predicts breast cancer from blood measurements https://www.nature.com/articles/s41388-018-0660-y 14
Predicts suicides https://www.ncbi.nlm.nih.gov/pubmed/30474411 15
Protein function given their aminoacid sequence https://www.nature.com/articles/s41598-017-03557-4 16
Predicts nanomaterial properties given structure https://www.nature.com/articles/s41524-017-0045-8 17
Just Add Data Bio Release this week! 18
Conclusions o Machine Learning is a very exciting field! o Research opportunities for intelligent, knowledgeable, ambitious, hard-working students and o Employment opportunities o Contact for more info: tsamard@csd.uoc.gr o Check out: jadbio.com 19
Recommend
More recommend
