Virtual screening with quantum computing! #Quantum_computing #chemoinformatics #memorandum

Quantum computing is one of the hot area in these days. Now google reported exciting article to nature.

Quantum computing reach is also very useful for drug discovery. Some days ago I found interesting article published by researcher in 1QBit, Univ. Drive, Accenture Labs and Biogen. URL is below.
A Quantum-Inspired Method for Three-Dimensional Ligand-Based Virtual Screening

The authors used quantum computer for Ligand Based Virtual Screening(LBVS). Their strategy is below.

Made molecular graph which like pharmacophore representation of molecule. And the search maximum common sub graph(MCS) with quantum computing.

Find MCS problem is NP hard. So it requires huge computational cost. But they solved the issue with quantum annealar.

To find the MCS, they used conflict graph. I’m not familiar the concept but regarding the publication, the graph made from two molecular graphs.

From the fig3 in the article. (Pls check original article because I can’t share figure)

Construction of the conflict graph G c from two given graphs G 1 and G 2 . A vertex (v 1 , v a ) is added to G c because at least one of the labels in the set of labels associated with v 1 from G 1 (3 v 1 ), and v a from G 2 (3 v a ) match. In this case, the set of labels match exactly so we designate the new vertex (v 1 , v a ) as an exact match. The rest of the vertices in the conflict graph are added in the same way. Edges are added according to two
conditions: bijective mapping and distance violations. Bijective mapping is violated if one of the nodes has been matched twice (represented by a red edge). Distance violation aims to incorporate 3D molecular information (represented by a green edge). An edge between two vertices (e.g.,
between (v 1 , v b ) and (v 2 , v c )) is added if the Euclidean distance between v 1 and v 2 is not comparable to the Euclidean distance between v b and v c .
Formally, an edge is added if |d(v 1 , v 2 ) − d(v b , v c )| > ε (ε = 0.4 in this example).

Regarding the concept described above, less edge graph is preferred for maximum common substructure.

And by using the method, graph based approach outperformed Morgan finger print based LBVS against several targets.

It indicates that quantum computer is useful for drug discovery.

Unfortunately the calculation performance of quantum computation is not described in this article.
I would like to know comparison between current traditional computation and quantum computation.