#No. 1 Weekly# Researchers such as David Baker, University of Washington (Seattle) developed a new method based on deep learning - ProteinMPNN (message passing neural network) to design proteins de novo.

Further, researchers combined with phantom (hallucination) protein design strategy starting from random sequences, and directly designed a protein multimer that can meet this parameter from topology parameters (number and length, etc.) . Moreover, these polymers diversity high , and many of them have different structures from known proteins in nature, thus greatly expanding the space and potential application range of protein de novo design (2).

or above work was published back-to-back in Science on September 15, 2022; corresponding author David Baker said that these deep learning-based strategies will gradually realize dynamic and functional protein design of (1–3).

Comment(s):

ProteinMPNN's successful protein design rate is high, and a strong universality is perhaps because it is more inclined to consider global interactions 's characteristics are similar to the actual protein folding principle of .

In addition, starting from the protein design, the massive design analyzes the sequence and protein structure (expected or otherwise) and its Properties (solubility, aggregation state, etc.) association may also obtain protein fold and its chemical properties new insights .

References:

1. J. Dauparas et al., Robust deep learning–based protein sequence design using ProteinMPNN. Science (80-. ).(2022), doi:10.1126/SCIENCE.ADD2187.

2. B. I. M. Wicky et al., Hallucinating symmetric protein assemblies. Science (80-. ). (2022), doi:10.1126/SCIENCE.ADD1964.

3. Beyond AlphaFold: AI excels at creating new proteins, (available at https://phys.org/news/2022-09-alphafold-ai-excels-proteins.html).

Original link:

https://www.science.org/doi/10.1126/science.add2187

https://www.science.org/doi/10.1126/science.add1964