Ornate 3D CNN

About

Oriented Routed Neural network with Automatic Typing is a method for protein quality assessment.

Method

Ornate is a residue-wise scoring method. It first constructs a three dimensional map representing the structure of the residue, and its neighborhood.

This representation is used as input of a deep neural network to predict the score of this residue

 

Reduced typing

Here, we provide an SVD decomposition of the retyper matrix. Each atom type is associated with a 15 dimensional feature vector. The components are ordered by importance, one can use all of them or only the N first components of each atom type. type_svd

Authors

Guillaume Pagès¹ , Benoit Charmettant^1,2& Sergei Grudinin¹,

¹Nano-D team, Inria/CNRS Grenoble, France

²Ecole CentraleSupélec

e-mail: Sergei.Grudinin @ inria.fr

 

Download

Ornate v. 0.1

References

Protein model quality assessment using 3D oriented convolutional neural networks. Guillaume Pagès, Benoit Charmettant, Sergei Grudinin. Bioinformatics, btz122, ⟨10.1093/bioinformatics/btz122⟩

What does deep learning see in 3D protein structures? Guillaume Pagès, Benoit Charmettant, Sergei Grudinin. Submitted.

License

All rights reserved. The academic version is free. Source code is available on demand.