Different version of Tophat gives me different number of mapped reads

Question

[BioDH]

I have paired-end sequenced RNA-seq files(Illumina, fastq).

I trimmed the reads by trimmomatic

java -jar <path>/trimmomatic-0.36.jar PE -threads 4 -phred33 1.fastq 2.fastq 1_trim1.fastq 1_unpaire    d1.fastq 2_trim2.fastq 2_unpaired2.fastq ILLUMINACLIP:<path>/TruSeq3-PE-2.fa:3:30:10 SLIDINGWINDOW:5:20 MINLEN:20

around 14M paired reads were survived.

I aligned the trimmed fastq files to genome by tophat(old version = 2.1.0, new version = 2.1.1) on exactly same argments.

# This is old version (2.1.0)  
 tophat --num-threads 4 --read-mismatches 1 --read-edit-dist 2 --read-realign-edit-dist 1000 -a 8 -m 0 -i 30 -I 1000 -g 1 --min-segment-intron 30 --max-segment-intron 1000 --segment-mismatches 1 --segment-length 25 --library-type fr-secondstrand --max-insertion-length 3 --max-deletion-length 3 --no-coverage-search -r 100 --mate-std-dev 20 -o ./local_tophat_old_alignments <genome> 1_trim1.fastq 2_trim2.fastq

 # This is new version (2.1.1)

~/script/tophat-2.1.1/tophat --num-threads 4 --read-mismatches 1 --read-edit-dist 2 --read-realign-edit-dist 1000 -a 8 -m 0 -i 30 -I 1000 -g 1 --min-segment-intron 30 --max-segment-intron 1000 --segment-mismatches 1 --segment-length 25 --library-type fr-secondstrand --max-insertion-length 3 --max-deletion-length 3 --no-coverage-search -r 100 --mate-std-dev 20 -o ./local_tophat_new_alignments <genome> 1_trim1.fastq 2_trim2.fastq

I check the number of reads

samtools veiw -c accepted_hits.bam

old version gave me 6,910,198

new version 27,645,322

I don't know why the number of reads are so different.

Just in case, I show you align_summary.txt

#old version
Left reads:
          Input     :   3540517
           Mapped   :   3449572 (97.4% of input)
Right reads:
          Input     :   3540517
           Mapped   :   3460626 (97.7% of input)
97.6% overall read mapping rate.

Aligned pairs:   3380776
                 1025738 (30.3%) are discordant alignments
66.5% concordant pair alignment rate.

#new version
Left reads:
          Input     :  14189364
           Mapped   :  13781744 (97.1% of input)
Right reads:
          Input     :  14189364
           Mapped   :  13863578 (97.7% of input)
97.4% overall read mapping rate.

Aligned pairs:  13503616
                 4100383 (30.4%) are discordant alignments
66.3% concordant pair alignment rate.

Could anyone please explain it?

[Joe]

All versions of TopHat are the old version. It should not be used any more - the authors themselves state this.

Answer 1

[swbarnes2]

Obviously the "old" run omitted a lot of of your reads. Figure out why. I'd start by just rerunning it again, in case it got randomly stopped mid run.

[h.mon]

swbarnes2's answer is the correct answer to your question: somehow the run with Tophat 2.1.0 didn't read the entire fastq files.

However, jrj.healey comment above is the correct answer to your needs. From the Tophat2 page:

Please note that TopHat has entered a low maintenance, low support stage as it is now largely superseded by HISAT2 which provides the same core functionality (i.e. spliced alignment of RNA-Seq reads), in a more accurate and much more efficient way.