Hi, I am Hojin Lee, a graduate student at Korea University's Jungmin Choi's laboratory.
First of all, thank you for your interesting research and great efforts.
I have been enjoying reading the three series of papers you have published. Currently, there is no pre-calculated score available, so I'm going to utilize the website first. As I understand the paper "The landscape of tolerated genetic variation in humans and primates", it is trained using 4.3 million common benign variants from 233 primate species, and then uses human protein structure information to predict the pathogenicity of possible point mutations.
However, the website only shows the primate AI 3D scores of the variants found in the primate 233 species. According to the paper, only about 30% of variants overlap with gnomAD, so I wondered if there was a plan to provide primate AI 3D scores for human private included in gnomAD. Of course, it would be better if I could obtain pre-calculated data on possible variants in the human genome, as requested below.
In the case of papers modeling a primate AI 3D, 0.821 is used as a cutoff to distinguish between benign and pathogenic, and in the case of papers dealing with PRS, 0.5 is used as a cutoff. Can you recommend which cutoff to use?
Also, I found a problem with primad web. When searching through the loci, such as 11-77181508-77181548, the primate species are not listed in the source section. The variant, 11-77181548-G-A, is an example, and the variant is originally generated from not only primates but also humans. I attached the example.
Dear Illumina,
Hi, I am Hojin Lee, a graduate student at Korea University's Jungmin Choi's laboratory.
First of all, thank you for your interesting research and great efforts. I have been enjoying reading the three series of papers you have published. Currently, there is no pre-calculated score available, so I'm going to utilize the website first. As I understand the paper "The landscape of tolerated genetic variation in humans and primates", it is trained using 4.3 million common benign variants from 233 primate species, and then uses human protein structure information to predict the pathogenicity of possible point mutations.
However, the website only shows the primate AI 3D scores of the variants found in the primate 233 species. According to the paper, only about 30% of variants overlap with gnomAD, so I wondered if there was a plan to provide primate AI 3D scores for human private included in gnomAD. Of course, it would be better if I could obtain pre-calculated data on possible variants in the human genome, as requested below.
In the case of papers modeling a primate AI 3D, 0.821 is used as a cutoff to distinguish between benign and pathogenic, and in the case of papers dealing with PRS, 0.5 is used as a cutoff. Can you recommend which cutoff to use?
Also, I found a problem with primad web. When searching through the loci, such as 11-77181508-77181548, the primate species are not listed in the source section. The variant, 11-77181548-G-A, is an example, and the variant is originally generated from not only primates but also humans. I attached the example.
Best, Hojin