The next step was to use AlphaFold AI-based protein folding software to examine how CRISPR complexes interact with them. The updated version is designed to handle interactions between proteins and nucleic acids, as well as complexes of multiple proteins. So the researchers gave them an AlphaFold version of the target DNA sequence, a guide RNA, a Cas9 sequence, and an enzyme that could chemically modify the bases and attach it to Cas9.
Unfortunately, it choked and I clearly put one of the proteins in the wrong place.
Still, the team kept things simple, feeding AlphaFold only with DNA, RNA, and Cas9 protein. This is because the latter is the main factor determining its sequence specificity. This worked much better, producing structures that matched those determined by experiments with real nucleic acids and proteins.
The researchers discovered a general pattern by comparing the structures of AlphaFold produced when feeding different on- and off-target sites. Due to many of the off-target sites (approximately two-thirds), the Cas9 protein adopted a slightly different structure. However, almost all of them (more than 95%) changed which amino acids contact the RNA. Therefore, even though Cas9 maintains its normal structure, there are clearly some cases in which the amino acids within it change in a way that corresponds to incorrect bases at off-target sites.
Conveniently, AlphaFold is already set up to identify so-called “contact probabilities,” or the probability that any two items, such as amino acids or nucleotides, will be within a very short distance (8 angstroms). Researchers can capture and compare the output of contact probability analysis for on-target and off-target sites to pinpoint which amino acids in Cas9 modify contacts when there is a mismatch between guide RNA and DNA. They named this computerized analysis setup “ContactSeek.”
better targeting
ContactSeek itself tended to generate a large list of amino acids that migrated upon binding to off-target sites. The researchers then focused on regions of the Cas9 protein where these amino acids clustered, and considered these regions to be signs of adaptation to differences caused by base mismatches. They then began testing versions of Cas9 with different amino acids at these sites.
