Quick optmaven: an efficient computational framework for the de novo design of fully human monoclonal antibodies

Open Access
Allan, Matthew Frederick
Area of Honors:
Chemical Engineering
Bachelor of Science
Document Type:
Thesis Supervisors:
  • Costas D Maranas, Thesis Supervisor
  • Andrew Zydney, Honors Advisor
  • monoclonal antibody
  • mAb
  • mAbs
  • protein design
  • antibody design
  • protein engineering
  • antibody engineering
  • antibody development
  • immunology
  • computational biology
  • bioinformatics
  • IPRO
  • OptMAVEn
  • OptCDR
  • pharmaceuticals
  • biologics
Monoclonal antibodies (mAbs) have become one of the most promising classes of therapeutics. Currently, dozens of mAbs are FDA-approved for autoimmune disorders, cancers, and infectious diseases. While there exist robust laboratory methods for designing mAbs for new antigens, these expensive and time-consuming (i.e. 3 – 6 months) methods are incapable of developing therapeutic antibodies rapidly during an epidemic. Potentially, computational methods of engineering mAbs could expedite this process. The first software capable of designing antibody variable domains de novo—OptMAVEn—was published in 2014. Despite designing three mAbs that bound a dodecapeptide antigen, OptMAVEn proved too time- and storage-intensive to design mAbs for Zika during the 2015 – 2016 epidemic. Here, we present Quick OptMAVEn, a new implementation of OptMAVEn. We show that Quick OptMAVEn can design variable domains of equivalent quality in 74% less time using 84% less disk storage, relative to OptMAVEn. Furthermore, we have used Quick OptMAVEn to design 50 antibodies for Zika, 9 of which are predicted to bind the antigen with greater affinity than an antibody isolated from a human patient. Quick OptMAVEn achieves better performance by using more efficient algorithms, more compact representations of antigen structures, and a novel k-means clustering step. We plan to create a web server to share Quick OptMAVEn with other labs.