A divide and conquer approach (DACA) to predict high fidelity structure of large multidomain protein BRWD1
Therapeutic importance in inhibiting Bromodomain and WD Repeat Domain containing BRWD1 against numerous human pathophysiological processes including cancers prompts prediction of a workable structure of this large protein. Here, a novel divide and conquer strategy was adopted to utilize smaller overlapping sequence-fragments of BRWD1 to further utilize their predicted structures as derived templates for prediction of complete BRWD1 structure in absence of its desired homologues in the template database. The novelty of this methodology stemmed from the requirement of templates of high sequence similarity in any comparative model based predictors whereas, the own fragments of the same target protein, BRWD1 could successfully fulfill this criteria. Additionally, the outputs of different high performing predictors including AlphaFold and RoseTTAFold were systematically integrated under the premise of Inductive Reasoning. The resultant structures were validated using existing validation parameters. Finally, a new validation paradigm was adopted to screen the best structure from the result presenting in-silico studies of known interactions of BRWD1 with various small molecules like, BD inhibitors, modified histone tails, DNA motifs and interacting proteins. The algorithm proposed in this work also paved the way for prediction of authentic structures of large size proteins.