cufflinks does not understand strands?

[staszekdh]

Dear All,

I used STAR to map paired-end reads onto the reference genome:

~/apps/STAR/bin/Linux_x86_64/STAR --runMode alignReads \
				  --runThreadN 10 \
				  --genomeDir ../../../genome/GCA_..../ \
				  --readFilesIn R1.fastq R2.fastq \
				  --outFilterIntronMotifs RemoveNoncanonical \
				  --outSAMattrIHstart 0 \
				  --alignIntronMax 1 \
				  --alignMatesGapMax 200 \
				  --outFileNamePrefix ./out/ \
				  --limitBAMsortRAM 1048477838 \
				  --peOverlapNbasesMin 0 \
				  --quantMode GeneCounts \

The mapping looks good:



Note that reads are colored according to the "first-of-pair strand" rule.

The next step is reference-guided transcripts reconstruction. To this end, I tried to use cufflinks, but without success:

~/apps/cufflinks-2.2.1.Linux_x86_64/cufflinks -o ./out/ \
					      -p 10 \
					      --library-type fr-firststrand \
					      -g ../../../genome/GCA_XXX/GCA_XXX.gff \
					      ../star/out_mock_rep2_clean/Aligned.sortedByCoord.out.bam

First, I got a weird warning:

Warning: Using default Gaussian distribution due to insufficient paired-end reads in open ranges.  It is recommended that correct parameters (--frag-len-mean and --frag-len-std-dev) be provided.

Second, the predicted transcripts look odd. For example, the two genes (gene0 and gene1 - see image above) were merged into a single transcript, despite they have the opposite orientation!

I would appreciate some hints on that.

Best wishes,
Staszek

[staszekdh]

Problem solved. I used HISAT2 with "--rna-strandness RF" + stringtie and now I get reasonable results.

HISAT2:

$HISAT2_HOME/hisat2 -q \
                -p 10 \
                -x ../../../genome/genome/XX \
                -1 ../B1R_S1_R1_001.clean.fastq \
                -2 ../B1R_S1_R2_001.clean.fastq \
                -S out_mock_rep1_clean/alns.sam \
                --fr \
                --rna-strandness RF \
                --dta

stringtie:

~/apps/stringtie-1.3.5.Linux_x86_64/stringtie \
	../hisat/out_mock_rep1_clean/alns.sorted.bam \
	-o ./out_mock_rep1_clean/out.gtf \
	-p 10 \
	--fr \
        -G ../../../genome/xxx.gff

BTW: I found this post useful

[ashishbansal]

Tophat/Cufflinks are not for assembly, they are for alignment to an already assembled genome or transcriptome. What are you aligning your reads to? Also, if you have strand specific data, you shouldn't choose an unstranded library type. You should choose the proper one based on your library preparation method. The XS tag will only be placed on split reads if you choose an unstranded library type.