Take the finished dish,
or take the ingredients?
Open four public expression datasets and the value ranges are all over the place β one runs to 96,841, another starts at β8.3. Each depositor processed theirs differently, and those decisions cannot be undone.
What existed before those numbers is in SRA: raw sequence fragments, 46 million experiments of them.
Over these 20 lessons we pull the real metadata for 713 runs and look at what it says β eight different library selections inside RNA-Seq alone, a 4,828-fold spread in depth, and a Disease field that is 100% empty. Python runs right inside this page.
How you'll learn
Fourteen columns that matter out of forty-seven β from 713 runs actually retrieved.
PolyA selection or ribosomal depletion β they capture different RNA.
All 713 of them, in data found by searching for ALS. So where are the labels? Lessons 9 and 10.
4.4 terabytes, 22 days, and twelve decisions you must write down. Go down only when you have a reason.