Reading the experiment
nature already ran
Once you have picked a target, there is one way to ask a human whether you were right: do people whose version of this gene is slightly different get the disease less often? Everyone's genes differ a little, and which allele you got was decided at random at conception โ nature has already run the experiment.
Targets with that kind of evidence are said to be twice as likely to reach approval. But what exactly is that support?
Over these 20 lessons we download and take apart 1.19 million GWAS Catalog associations and TCGA somatic mutations: a median odds ratio of 1.14, the 92.7% of hits that change no protein, and the illusion that makes TTN look more mutated than KRAS. Python runs right inside this page.
How you'll learn
The whole catalog, downloaded and counted: median odds ratio 1.14, only 7.3% change a protein.
Linkage disequilibrium, the MAPPED_GENE trap, and a column that mixes odds ratios with betas.
TTN is mutated more often than KRAS in lung cancer. Divide by length and it sits at 1/163rd of TP53.
Hotspots mean inhibit; truncations mean there is nothing left to inhibit. Get the direction wrong and you make things worse.