Researchers used Pinal to design several types of functional proteins, relying on enough structural information to make the training process work. Their approach generated large numbers of candidates, filtered them computationally, and then tested dozens of designs. Only a small fraction worked, but the researchers say that success rate understates the method’s potential because many likely candidates were excluded for being too similar to known proteins.
Functional proteins
To test Pinal, the researchers asked it to design four substantially different proteins: a fluorescent protein, two enzymes, and a protein involved in a large multiprotein complex. The details varied between these cases, but each followed a similar process. Fluorescent proteins provide an example.
A classic fluorescent protein is GFP, or green fluorescent protein, whose discovery earned its discoverer a Nobel Prize. The researchers asked Pinal to produce 50,000 proteins that would fluoresce like GFP. They then filtered this large set using several criteria.
First, they rejected sequences that were too similar to known fluorescent proteins. At the same time, they selected for important features found in GFP, including the ability to form a barrel-like structure and the presence of key amino acids that can be chemically modified to produce fluorescence.
“We wanted something a little bit like GFP in general, but not too much like it in terms of specific amino acid sequences,” the researchers said. The filtering produced a prioritized list of just 29 candidate proteins. The candidates were tested in Escherichia coli, and one of them fluoresced.
The other examples followed a similar pattern: The researchers began with a large list, narrowed it computationally, and then tested dozens of designs. Only a few of the tested proteins worked.
That may appear to be a very low success rate, but the researchers caution against interpreting it that way. Many of the most likely effective proposals were similar to known GFP proteins and were therefore excluded. They also did not test candidates that failed to form structures resembling GFP, meaning that a broad range of potentially functional alternative fluorescent proteins may have been rejected before testing.
Source: arstechnica.com
