Mixed Experimental and AI-based Computational Studies of Edited Dystrophin STRs


Topic:

Pre-Clinical Research

Poster Number: 130

Author(s):

Nick Menhart, PhD, Illinois Institute of Technology, Jingwen Zhang, Vlad Stojkovic

Many therapies, current and envisioned, for DMD produce edited dystrophins. Most of these edits occur in the STR region which forms the bulk of this protein. This includes exon skipping therapy, as well as many gene-therapy techniques being developed that rely on edited mini-dystrophins, mostly at STR boundaries. These exon, or STR boundary, edits are the most studied edit types so far.
However, understanding of the consequences of these edits is incomplete. Evidence from BMD natural-history studies demonstrate that disease severity is linked to the nature of the edit. However, the high complexity of underlying patient defects, and heterogeneity of disease progression, mean we do fully understand what makes a “good” edit. It is accepted that any dystrophin is better than no dystrophin, but also widely believed that a full robust therapy will require a fully functional protein, with minimally damaging edits. Experimental studies of edited proteins have shown that they vary widely in stability and other properties, contributing to this complexity. On the horizon, CRISPR-based gene editing techniques such as CINDEL are more flexible, and able to produce a much wider range of edits, possibly allowing for the production of “better” = “more functional” edited proteins. However, is not immediately obvious exactly how to take advantage of this increased flexibility to produce “optimal” edits.
To address this, we have embarked upon a mixed computational/experimental study to screen such edits in silico by AI-based de novo protein structure prediction (Robetta, Alphafold) followed by molecular dynamic simulation to more rapidly identify putative optimal edits. This is followed by experimental confirmation of targeted experimental studies of selected edits to validate concordance of in silico prediction to real world properties. We will present studied on our first suite of CINDEL type edits, targeting the exon 51-59 region.