Why I built this

What do you get out of this?

In one line — you retrace, from the first step to the final validation, how Prof. Kwang-Hyun Cho's KAIST team broke the ‘immune desert’ of lung cancer and found DDX54.

The forty courses are not an arbitrary list. They are the parts you actually need in order to reproduce one discovery end to end, laid out in the order they get used. That process follows the classic arc of systems biology: data integration → computational simulation → molecular mechanism → validation in a living body. The four stages below are that arc, and the courses under each stage are where you do that stage by hand.

Multi-omics data integration · defining the patient group

📊 Data

TCGA genomic analysis — compute the tumour mutational burden (TMB) for lung adenocarcinoma (LUAD) patients and keep only those above the FDA's threshold of 10 mutations/Mb.
SBU-TIL AI matching — attach the AI-derived value from the same patient's H&E tissue image (the tumour-infiltrating lymphocyte fraction) to the genomic data.
Noise filtering — drop samples with defects in the antigen presentation pathway. In those, the immune system could never have recognised the tumour in the first place.

What this stage concludes — What remains is the roughly 30% of patients who are ‘TMB-high and immune-desert’ — loaded with mutations, yet with no lymphocytes. They are the target.

Courses where you do this by hand — Bioinfo Datasets, TIL, Bulk RNA-seq, Bioinformatics

Computational network modelling — finding the master regulator

💻 Computation

ARACNe — build a gene regulatory network (GRN) from lung cancer transcriptomes. It is a map of who switches whom on and off.
VIPER — on that map, compute the pivotal transcription factors that govern the immune-desert state. This is inference of an activity nobody measured.
DIGGIT — work back up to the topmost master gene pulling those factors' strings.

What this stage concludes — DDX54 is named, for the first time anywhere, as the gene commanding the whole desertification of the immune landscape — a name that came out of computation, not a bench.

Courses where you do this by hand — ARACNe, VIPER, DIGGIT, Gene Hunt, Digital Twin · foundations — Linear Algebra, Algorithms, Data Structures, Discrete Math

In vitro cell experiments — mapping the molecular control circuit

🔬 Mechanism

The microRNA mechanism — knocking DDX54 down in cancer cells raises Drosha and Dicer, and tumour-suppressing microRNAs such as miR-34b-5p come back to life.
The signalling pathways — conversely, when DDX54 is on, the MYC, JAK-STAT3 and NF-κB pathways switch on with it.
The immune-evasion proteins — riding those pathways, expression of CD38 and CD47, which bar immune cells from entering, is shown to be induced.

What this stage concludes — It becomes clear through which circuit the name stage 2 pointed at creates the desert.

Courses where you do this by hand — miRNA-mRNA, Gene Hunt, Bioinformatics

In vivo animal experiments and cross-validation — proving the therapeutic effect

🐁 Validation

Syngeneic mouse models — build models implanted with immune-resistant lung cancer cell lines (LLC1, KLN205).
Combination synergy — with DDX54 inhibition plus anti-PD1, tumours shrink dramatically and survival reaches 100%.
Spatial analysis (Xenium and single cell) — CD8+ T cells and M1 macrophages really do push all the way into the tumour core, confirming that the immune landscape has been remodelled from desert to inflamed.

What this stage concludes — Proof that the target computation pointed at works in a living body.

Courses where you do this by hand — Spatial, Immune Map, TIL

And the ground all of this runs on

All four stages above ultimately run on one server. You need somewhere to park tens of gigabytes of raw data, somewhere that will not fall over while it grinds for hours, somewhere that will not leak somebody else's patient data. The first courses are that ground — operating systems · computer architecture · computer networks · databases · security. They sit here in the same place so that nobody who came to learn analysis gets stuck in front of a server.

So what do you come away with?

You learn the route, not the name. Rather than memorising that "DDX54 matters", you walk the path that arrives at that name from a standing start.
You build the algorithms whole. Instead of importing ARACNe, VIPER and DIGGIT, you learn what they compute and write them by hand.
Nothing to install. Every line of code and every shell command runs right in the page. Nothing to prepare, nothing to pay.
Twenty-one courses on one account. Progress and scores collect in one place.

To be straight with you — the wet-lab work in stages 3 and 4 cannot be done in a browser. What you learn instead is how to read the data those experiments left behind and check by hand whether the same conclusion falls out. Stages 1 and 2 you can follow from beginning to end, exactly, on public data.

Browse all seventy-three courses →