cancer type prediction based on RNASeq gene expression data?

Question

[J.F.Jiang]

Hi,

I have several RNASeq data from FFPE, which were collected from several patients with cancer of unknown primary.

I processed the RNASeq data using STAR or hisat2, and convert the bam to FPKM or FPKM_UQ gene expression matrix.

TULIP was applied to predict the cancer type based on the expression data.

However, unexpected errors went out, either with FPKM or FPKM_UQ values.

I am wondering is there any other popular tool that can be applied on the expression data or the raw data(bam)?

Best,
Junfeng

[Darked89]

Just to be sure: are all the metastatic samples collected from the same organ? As: secondary tumors found in lungs with unknown origin? Patients with the same sex?

[J.F.Jiang]

all the samples are not metastatic.

For TULIP, I downloaded a sample FPKM_UQ value from GDC, and encountered the same error report from the tool. I've contact the author, hope the bug can be fixed.