BioTuring and Teddy Thinh

BioTuring 06/27: Gene Enrichment Analysis.
Motivation: Comparing 2 gene lists $A = {[G_{n1}, S_{n1}], [G_{n2}, S_{n2}], …}$ & $B = {[G_{m1}, S_{m1}], [G_{m2}, S_{m2}], …}$ to extrapolate on the results (e.g cancer vs non-cancer?)
Method: Get a gene pool, preferably on the whole $G_i$ set.
Arange from A $\rightarrow$ B to form a pathway.
Something about Fold Change (FC) metric here I don’t rmb well, but basically $FC = \log_2(\frac{G_{ni}}{G_{mi}})$, if FC > 0 then it’s good for hypo about gene list A, < 0 then good for gene list B.
For all gene in pathway, record its cummulative sum on FC. This will form a graph, axis x is A $\rightarrow$ B, axis y is score.
Highest one is Enrichment Score.
$\implies$ New problem, how do we know the Stats Significance of this score. Run it 10000 times and select one with p-score < 0.05? Seems reasonable, but resources are limited. This is what Bioturing Algo team been researching for now, when assigning a score towards a gene is approximately even to every other gene, you can apply some guys’ work on distribution to get 5000x faster than the conventional approach.

Took an entrance test @ BioTuring then

---

Met Teddy Thinh:
Here’s my two cents about the conversation:

• Graduate education requires self-commitment and curiosity to learn new things, and applying it to the company at which you’re having internship.
• Grad program are for connection probably, at least in HCMUS.
• Prof. Ly Quoc Ngoc (Computer Vision) is goated, his mindset about research is something to be recognised by fellow researchers.