Contradiction in Harzard ratio and Log fold change

Question

[Amir.Asgary]

Hi there. I am working on a pan-cancer project that contains 12 cancers. The data in use is TCGA RNA-seq data. I have found 37 genes that are involved in survival of patients.I performed a Differential expression between normal and cancer patients. then I have performed a CoxPH model between Dead and Alive Cancer patients on selected genes. dead = 1 and survive= 0.

I calculated LogFC of each gene expression between cancer and normal.

I know that genes with positive LogFC and +1 HR are known as bad prognostic. also genes with negative LogFC and 0<HR<1 are known as good prognostic but my problem is that some genes contain negative Log FC and +1 HR or positive LogFC wirh HR value less than 1.

What does it mean? Thanks

Answer 1

[Kevin Blighe]

Hi,

There is no issue here, in my opinion.

Consider the following 3 scenarios:

• a tumour suppressor gene is deleted or heavily methylated, and its expression via RNA-seq is then virtually 0 (negative log [base 2] fold-change). This results in an increased risk of death (HR > 1).
• an oncogene 'driver' is amplified or has its expression increased via some other mechanism, and its expression via RNA-seq is then increased (positive log [base 2] fold-change). This also results in an increased risk of death (HR > 1).
• we observe increased B-cell and macrophage infiltration in the tumour nest, with corresponding increase in genes related to these, including CD20, MS4A1, and CD68 (positive log [base 2] fold change). This is protective (HR<1).

However, as with everything in cancer, there are 'double edged swords' in every which way that we look...

Kevin

[Amir.Asgary]

Thank you so much. I think so, that some signaling pathways are involved here specially some regulators like Tumor suppressors and oncogenes.