Task 1: Retrieve datasets by cancer type
Purpose: Find datasets for a specific cancer type.
How it works: The robot extracts disease names from your input and matches them against the dataset registry.
Best practice: Clearly mention the cancer name in your prompt (e.g., “prostate cancer”).
Prompt example: I need datasets related to prostate cancer.

---

Task 2: Find cell type proportion changes after treatment
Purpose: Ask whether a cell type increases or decreases after treatment.
How it works: The robot identifies the cell type and optionally a cancer context, then searches the cell proportion change table.
Best practice: Name one specific cell type (e.g., “macrophages”).
Prompt example: Does the proportion of B cells change after treatment?

---

Task 3: Explore gene expression dynamics
Purpose: Check treatment-induced expression changes of specific genes.
How it works: The robot extracts the gene name and looks up average log2 fold change values in the gene expression table.
Best practice: Specify the gene symbol (e.g., “EGFR”).
Prompt example: What is the expression change of EGFR during therapy?