Systems and methods for de novo design of protein interactions with learned surface fingerprints

Patent metadata

Field	Value
Publication number	US-2023395187-A1
Application number	US-202318206873-A
Country	US
Kind code	A1
Filing date	Jun 7, 2023
Priority date	Jun 7, 2022
Publication date	Dec 7, 2023
Grant date	—

Abstract

The present application relates to a computer-implemented systems and methods for protein interaction design using surface fingerprints. The method comprises predicting at least one target interface site with high binding propensity, wherein, optionally, the step of predicting at least one target buried interface site comprises generating at least one surface fingerprint associated with a protein interaction based on at least one protein interface, wherein the at least one surface fingerprint preferably embeds geometric and/or chemical features of molecular surfaces, identifying at least one binding seed that displays required features to engage the target site, and performing a binding seed transplantation to protein scaffolds to confer stability and additional contacts on the designed interface.

First claim

What is claimed is:

1 . A computer-implemented method for protein interaction design using surface fingerprints, the method comprising:
predicting at least one target interface site with high binding propensity; identifying at least one binding seed that displays required features to engage the target site; and performing a binding seed transplantation to protein scaffolds to confer stability and additional contacts on the designed interface.

2 . The computer-implemented method of claim 1 , wherein the at least one target interface site comprises at least one target buried interface site, and wherein predicting the target buried interface site comprises generating at least one surface fingerprint associated with a protein interaction based on at least one protein interface.

3 . The computer-implemented method of claim 2 , wherein the at least one surface fingerprint embeds geometric and/or chemical features of molecular surfaces.

4 . The computer-implemented method of claim 1 , further comprising performing a surface fingerprint-based search.

5 . The computer-implemented method of claim 1 , further comprising decomposing a protein molecular surface into overlapping radial patches, preferably with a radius of 12 Å.

6 . The computer-implemented method of claim 1 , further comprising computing a coordinate system in geodesic space.

7 . The computer-implemented method of claim 1 , further comprising using a first machine-learning model to output vector fingerprint descriptors that are complementary between interacting protein pairs and dissimilar between non-interacting pairs.

8 . The computer-implemented method of claim 1 , further comprising using a second machine-learning model to score matching surface patches, in particular for computing an interface post-alignment score.

9 . The computer-implemented method of claim 1 , wherein the at least one target interface site with high binding propensity is a prediction output by a machine-learning model trained on data comprising surface fingerprints and/or geometric or chemical features of molecular surfaces.

10 . The computer-implemented method of claim 1 , wherein the binding seed and/or the binding seed transplantation is used for one or more of:
obtaining a protein-based therapeutic; obtaining an antibody; obtaining an inhibitor; or obtaining a vaccine.

11 . A system comprising:
one or more processors, a memory; and computing instructions stored in the memory and configured to execute on the one or more processors, which when executed, causes the one or more processors to:
predict at least one target interface site with high binding propensity,
identifying at least one binding seed that displays required features to engage the target site, and
performing a binding seed transplantation to protein scaffolds to confer stability and additional contacts on the designed interface.

12 . The system of claim 11 , wherein the computing instructions are further configured, when executed by the one or more processors, to predict at least one target buried interface site comprises generating at least one surface fingerprint associated with a protein interaction based on at least one protein interface,

13 . The system of claim 12 , wherein the at least one surface fingerprint embeds geometric and/or chemical features of molecular surfaces.

14 . The system of claim 11 , wherein the computing instructions are further configured, when executed by the one or more processors, to perform a surface fingerprint-based search.

15 . The system of claim 11 , wherein the computing instructions are further configured, when executed by the one or more processors, to decompose a protein molecular surface into overlapping radial patches, preferably with a radius of 12 Å.

16 . The system of claim 11 , wherein the computing instructions are further configured, when executed by the one or more processors, to compute a coordinate system in geodesic space.

17 . The system of claim 11 , wherein the computing instructions are further configured, when executed by the one or more processors, to execute a first machine-learning model to output vector fingerprint descriptors that are complementary between interacting protein pairs and dissimilar between non-interacting pairs.

18 . The system of claim 11 , wherein the computing instructions are further configured, when executed by the one or more processors, to execute a second machine-learning model to score matching surface patches, in particular for computing an interface post-alignment score.

19 . The system of claim 11 , wherein the at least one target interface site with high binding propensity is a prediction output by a machine-learning model trained on data comprising surface fingerprints and/or geometric or chemical features of molecular surfaces.

20 . A computer-readable medium storing a computer program, the computer program comprising instructions which, when the program is executed by one or more processors, cause the one or more processors to:
predict at least one target interface site with high binding propensity; identifying at least one binding seed that displays required features to engage the target site; and performing a binding seed transplantation to protein scaffolds to confer stability and additional contacts on the designed interface.

Assignees

• Ecole Polytechnique Fed Lausanne Epfl
• Imperial College Innovation Ltd

Inventors

• Gainza Cirauqui Pablo
• Wehrle Sarah
• Van Hall-Beauvais Alexandra
• Marchand Anthony
• Scheck Andreas
• Harteveld Zander
• Sverrisson Freyr
• Bronstein Michael
• Ferreira De Sousa Correia Bruno Emanuel

Classifications

• G16B15/30 Primary

  Drug targeting using structural data; Docking or binding prediction · CPC title

• G16B20/30

  Detection of binding sites or motifs · CPC title

• G16B40/20

  Supervised data analysis · CPC title

• G16B15/20

  Protein or domain folding · CPC title

• G16B20/50

  Mutagenesis · CPC title