Hisat2 vs Tophat2 variation between tech reps

kuristy · 10-03-2017, 09:28 AM

I have a question regarding the variability of FPKM I am seeing between technical replicates in the Hisat2, stringtie pipeline. When I plot technical replicates by log2 FPKM, I am getting Spearman correlations of 0.6. I used the same fastq files in the tophat, cufflinks pipeline (default settings) and had Spearman correlations of 0.9. I am wondering which parameters I can change in Hisat2 to make it behave more like tophat2. In hisat, I have tried several different parameters and I am getting similar results regardless of the parameters I use.

The parameters I have tried in hisat2 are as follows:

-k 2 --max-seeds 15 --no-mixed

-k 3 --max-seeds 10

-k 10 --max-seeds 10

default settings

I am getting a high percentage of multi-mapping (~20%) and I am wondering if that is a problem.

Thank you for your help!

kuristy · 10-04-2017, 02:23 PM

I have answered my own question, but I am posting just in case someone else has a similar issue.

I ran the same tophat data with stringtie and got a similarly bad correlation between technical replicates of 0.62. Now I know that the problem must be with stringtie and not with hisat2.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Uneven number of biological reps on DESeq	sulfatereducers	Introductions	2	11-07-2016 01:58 PM
samtools flagstat issues w/ star, hisat2, gsnap & tophat2	jmartin	Bioinformatics	6	09-30-2016 03:26 PM
HISAT2 vs. TopHat2: Discrepancies between Hg19 & Hg38	Sbamo	RNA Sequencing	0	01-26-2016 06:05 AM
reps and analysis	Vale16S	Metagenomics	5	07-30-2014 08:48 AM
Statistical treatment of biological reps from pooled samples	danwiththeplan	RNA Sequencing	7	10-23-2013 09:28 PM

10-03-2017, 09:28 AM	#1
kuristy Junior Member Location: Florida Join Date: Oct 2017 Posts: 3	Hisat2 vs Tophat2 variation between tech reps I have a question regarding the variability of FPKM I am seeing between technical replicates in the Hisat2, stringtie pipeline. When I plot technical replicates by log2 FPKM, I am getting Spearman correlations of 0.6. I used the same fastq files in the tophat, cufflinks pipeline (default settings) and had Spearman correlations of 0.9. I am wondering which parameters I can change in Hisat2 to make it behave more like tophat2. In hisat, I have tried several different parameters and I am getting similar results regardless of the parameters I use. The parameters I have tried in hisat2 are as follows: -k 2 --max-seeds 15 --no-mixed -k 3 --max-seeds 10 -k 10 --max-seeds 10 default settings I am getting a high percentage of multi-mapping (~20%) and I am wondering if that is a problem. Thank you for your help!

10-04-2017, 02:23 PM	#2
kuristy Junior Member Location: Florida Join Date: Oct 2017 Posts: 3	I have answered my own question, but I am posting just in case someone else has a similar issue. I ran the same tophat data with stringtie and got a similarly bad correlation between technical replicates of 0.62. Now I know that the problem must be with stringtie and not with hisat2.