What is it about?
A recurrent neural network with LSTM cells was trained on known helical antimicrobial peptides (AMP) to generate new examples that lie close to the known AMP sequences in peptide space.
Featured Image
Why is it important?
It is the first application of generative recurrent neural networks for the de novo design of peptide sequences. The model is not restricted to antimicrobial peptides but can be applied to any amino acid sequence of interest to design novel but closely related examples for biological testing.
Perspectives
I hope that this article opens a new field for peptide design solely based on amino acid sequences without the need for molecular descriptors. It can be compared to the way people read text and learn grammar. The learned "peptide grammar" can then be used to construct new peptide sequences, which enables the discovery of new bioactive molecules in many research fields.
Mr Alex T Müller
ETH Zürich, D-CHAB
Read the Original
This page is a summary of: Recurrent Neural Network Model for Constructive Peptide Design, Journal of Chemical Information and Computer Sciences, January 2018, American Chemical Society (ACS),
DOI: 10.1021/acs.jcim.7b00414.
You can read the full text:
Contributors
The following have contributed to this page
