Dicle Hasdemir – Elsevier

Dicle Hasdemir – Elsevier

Dicle is a Data Scientist in Content & Innovation (C&I) department at Elsevier. Content & Innovation (C&I) team combines NLP and ML methods with domain expertise in order to enrich content into data structures.

Dicle has recently joined Elsevier and is working on interdisciplinary projects at the intersection of modeling, life sciences and data science. 

Prior to joining Elsevier,  she worked on algorithm design for biological data integration and application of machine learning methods on molecular/medical data. Previously during her PhD in bioinformatics, she focused on designing proper testing and validation strategies for different types of systems biology models.

Title

Sequential Literature-informed Refinement of Mechanistic Hypotheses in Biology and Medicine

Abstract

Natural Language Processing now make it possible to extract statements about mechanistic relationships directly from the research community’s understanding of biological processes. By integrating experimental and clinical data with these statements, we can model these processes as logical decision-making circuits. Analysis of the differences in competing candidate models helps identify maximally informative experiments. Recently we developed a model of immune response to upper respiratory infection based on regulatory mechanisms documented in over 20,000 journal publications. We identified 35 candidate models consistent with available data. However, different behaviors were predicted from one model to the next.  Consensus among the family of competing models identified the immune mediators IL-4 and CXCL8 as maximally informative measurements. By performing these measurements, the set of feasible models was reduced from 35 to 11. Casting existing data in the context of documented regulatory mechanisms efficiently focuses experimentation in an advance over conventional design of experiments.