Work

Prediction of crucial interactions for icosahedral capsid self-assembly by energy landscape atlasing

Year: 2020
Type: preprint
Abstract: We predict crucial inter-atomic interactions between monomers for the autonomous assembly of three icosahedral viral capsids, Adeno Associated Virus type 2 (AAV2), Murine Parvo Virus (MVM), and Bromo Mosaic Virus (BMV), by approximating changes in the partition function of the capsid assembly landscape when an interaction is removed. At the interface-scale, the partition function computation uses \emph{atlases} of selected subassembly landscapes for each interface type and their abbreviated \emph{bar-codes}. The first (direct) prediction of cruciality of an interaction for successful capsid assembly assumed the interface assembly is essential for capsid assembly. The interface-scale prediction was blind to the mutagenesis data used for validation. The atlases are generated by the geometric method and curated opensource software EASAL (efficient atlasing and search of assembly landscapes), whose input consists of (a) the monomer geometry - atom coordinates; and for each interface type, (b) pair potentials for a candidate set of interactions along with Van der Waals sterics, and (c) structure of selected, small subassemblies. The method comes with rigorous accuracy-efficiency trade-off estimates. A second two-scale measure of cruciality of interactions and residues for capsid assembly is obtained as follows. Combinatorial entropy is used to define the cruciality of an interface for successful assembly of the capsid. This is combined with the interface-scale prediction using statistical learning. The learning algorithm uses - for training - a fraction of the mutagenesis data, with the remainder of the data being used for validation. By reliably and rapidly narrowing down target interactions, these in-silico predictions can inform and reduce the time spent on time-consuming in-vitro and in-vivo experimentation.
Language: English
Open Access status: green
OpenAlex ID: W3182038925
Cites: 70
Cited by:
Related to: 10