(1) Generate a scoring system for a set of published genes (Cox scores available) from gene expression data (fold changes) from a cohort of patients.
(2) Test is the scores associate with some defined clinical variables (e.g. disease severity, medication use, etc.)
Checking for insights on best method/strategies, especially to get feedback on my initial thoughts/plan outlined below:
- Use respective Cox scores to select/narrow gene list from the published literature, or from weighted scores for each gene (Cox.scores_geneA)x(Fold.changes_geneA) per patient?
- Determine putative signature based on statistical significance (? paired t-tests healthy vs diseased subjects’ gene expression data)
- Test whether the defined signature (or its composite score; Cox or weighted) correlates with specific clinical variables like medication use.
Questions: Are there established approaches in this case? Have you done similar analyses before?
P.S.- For each gene in the published set, we have the Cox score (from published literature) and fold changes (gene expression data), but we do not know how best to use this data along with the to derive and test plausible signatures in our cohort.
Thank you very much!