[Student Seminar] 12/18/18 Israel Desta + Christos Michas

     Accurate determination of antibody-antigen interactions is an important step in several academic and pharmaceutical endeavors. Experimental structure determination has not been able to catch up with identifying biological protein complexes for drug discovery, as it is too costly and time consuming. With computational methods such as protein docking, this gap can be potentially bridged. One of the most widely used protein-docking servers is ClusPro developed in our lab in 2004. On the latest benchmark set of proteins, ClusPro was able to predict 48% of antibody-antigen complexes accurately. The goal of my PhD is to improve on ClusPro’s prediction accuracy of antibody-antigen complexes. I will be using a two-pronged approach in order to achieve this. Firstly, I will be improving the contact-based potential called Decoys as a Reference (DARS) that is currently in use in ClusPro. The antibody specific DARS (aADARS), currently utilizes only 4 of the known 18 atom types and also uses 1 distance bin as a measure of contact. I am working on expanding both the number of atom types considered and also the number of distance bins. Secondly, I am planning to use ClusPro’s capabilities in predicting true epitope as an additional screening of predicted poses. Preliminary results show that ClusPro predicts true epitopes the majority of the time despite getting the actual structure of the complex wrong. The latter approach will need to utilize conventional machine learning and/or deep learning to enhance accuracy.

     The cardiac cycle is divided into four distinct phases, during which the myocardium is exposed to different patterns of blood pressure and tissue elongation. The pressure-sensitive cardiac valves give rise to these four phases, and changes in the differential pressure and elongation of each phase impacts cardiac tissue structure and function. The absence of a valvular mechanism in current in vitro cardiac micromodels limits the ability of these models to simulate the in vivo ventricular function and to translate their findings to cardiac output metrics traditionally used in the clinic. Using two-photon direct laser writing, we fabricate a passive microfluidic check valve that can impose unidirectional, pulsatile flow when exposed to oscillatory fluid pressure. We demonstrate opening and closing pressures below 1Pa, rendering our construct suitable as a miniaturized valve replica in an in vitro cardiac system. Future work aims to use this technology to mimic all phases of the cardiac cycle and study the cardiac remodeling that occurs when each phase is altered.