Using a gene signature to stratify patient samples

Question

[salvatore.raieli2]

Hi,

I have mRNA expression data and through machine learning analysis I identified a gene signature. I have a gene list with relative weights. I want to stratify patient in two group, one group high enriched in the signature and one low enriched. I would like to assign a signature score to each patient sample.

I thought about calculate the z-score for each gene in the signature and calculate the average for each patient. I also thought about run a PCA (or t-SNE) on the patient samples (only taking in account as variables the genes in the signature) and then using a cluster algorithm.

But I am not convinced with these approaches. Do you have some suggestions on how I can calculate a signature score for each patient? There some R libraries that can you suggest?

Thank you for the help or any suggestion

[user_without_id]

How many samples per patient?

Answer 1

[Kevin Blighe]

Care Salvatore,

Alcuni pensamenti / Some of my thoughts: A: What is the best way to combine machine learning algorithms for feature selectio

I have a gene list with relative weights.

If you have weights, then fit a Bayesian logistic regression model with the weights as priors.

A dopo,

Kevin