Google DeepMind's AI successor predicts how 71M mutations cause disease(endpts.com)
endpts.com
Google DeepMind's AI successor predicts how 71M mutations cause disease
https://endpts.com/google-deepminds-alphafold-successor-predicts-how-71m-mutations-cause-disease/
7 comments
Thanks for this...
% Mutated genes with at least 1 allele classified as benign: 95.72% % Mutated genes with at least 1 allele classified as ambiguous: 2.71% % Mutated genes with at least 1 allele classified as pathogenic: 1.57% % Mutated genes with 2 alleles classified as benign: 38.37% % Mutated genes with 2 alleles classified as ambiguous: 0.54% % Mutated genes with 2 alleles classified as pathogenic: 0.39%
The first one I checked freaked me out a bit (I have the AA mutation):
MTHFR C677T (Rs1801133). This mutation (the A allele) is associated with reduced enzyme activity, elevated total homocysteine levels and altered distribution of folate [1]. People with an “A” allele for this mutation present a 35% decrease of the normal enzyme activity and “AA” individuals a 70% decrease...
% Mutated genes with at least 1 allele classified as benign: 95.72% % Mutated genes with at least 1 allele classified as ambiguous: 2.71% % Mutated genes with at least 1 allele classified as pathogenic: 1.57% % Mutated genes with 2 alleles classified as benign: 38.37% % Mutated genes with 2 alleles classified as ambiguous: 0.54% % Mutated genes with 2 alleles classified as pathogenic: 0.39%
The first one I checked freaked me out a bit (I have the AA mutation):
MTHFR C677T (Rs1801133). This mutation (the A allele) is associated with reduced enzyme activity, elevated total homocysteine levels and altered distribution of folate [1]. People with an “A” allele for this mutation present a 35% decrease of the normal enzyme activity and “AA” individuals a 70% decrease...
Frankly once you go down that rabbit hole you have to make sure that you aren't prone to anxiety. I wrote a tool (not sure if open-sourced) that just crosses the output of 23andme with dbSNP and the number of possible diseases that came up was impressive.
I did have the SIRT1 mutation associated with a long life so there's that.
I did have the SIRT1 mutation associated with a long life so there's that.
My understanding is that lot of 23andme data is driven by self-reported data. I wonder how much this would change if it got vetted by a medical professional.
Also kind of curious which percentage of the pathogenic mutated genes are recessive vs dominant.
Fun indeed!
Also kind of curious which percentage of the pathogenic mutated genes are recessive vs dominant.
Fun indeed!
Paper: https://www.science.org/doi/10.1126/science.adg7492?adobe_mc...
Non-paywall wired article: https://www.wired.co.uk/article/deepmind-ai-alphamissense-ge...
Non-paywall wired article: https://www.wired.co.uk/article/deepmind-ai-alphamissense-ge...
This will be 100x better before the first treatment hits human trials.
Imagine how powerful these systems would be with all of the health data the world has in paper vaults.
These are crazy times.
These are crazy times.
Usually experts won’t make educated guesses, but: does it beat those experts?
% Mutated genes: 1.97%
% Mutated genes that are benign: 93.98%
% Mutated genes that are ambiguous: 3.94%
% Mutated genes that are pathogenic: 2.08%
Does not really mean anything but my tired brain thought it was fun.