Reconstructing the genes' web
whole, out of the data
ARACNe is an algorithm that looks only at expression data and draws the web of which gene is ordering which about. Rather than taking that web from someone else's program, you build it twenty lines at a time, with your own hands — from mutual information all the way to the data processing inequality.
And you mark your own work against a simulated cohort whose answer is known. Python runs right inside this page.
How you'll learn
Mutual information is how many questions twenty questions just saved you, the DPI is tracing a rumour back to its source, a regulon is the troops a commander has under them. Meaning comes before the formula.
No libraries are called. You start at a single line of entropy, and at the end of lesson 20 a complete ARACNe is left in your hands.
You work on a simulated cohort whose true network is known, so at every step you can see what your web scores.
Every lesson has a laboratory widget — turn the bin count, the threshold and the DPI tolerance yourself and watch the result wobble.