Directed graphs abstracting biological entities and their interactions capture the systematic biomolecular function and allow systems biologists to simulate systems-level behavior. However, the construction of such graphs or networks is cumbersome and requires expert literature review. Automated natural language processing approaches can be employed to overcome this hurdle. The currently available techniques are limited in discovering novel interactions from literature as well as in building reliable models. We propose to combine state-of-the-art machine reading techniques with biological interactions present in online molecular databases. The outcomes from this approach can help provide literature-supported and biologically viable interactions for onward inclusion into network models. In addition, we propose to use meta-heuristic techniques to optimize these networks so that simulation results match experimental data or known biological behavior. The resulting pipeline can produce simulation-ready and biologically plausible network models in a semi-automated manner.

---

Committee: Dr. Safee Ullah Chaudhary (Supervisor), Dr. Basit Shafiq (Examiner)