[Nicolas]

Hi,
It would work the same if your File.mapped.bam has been aligned to a transcriptome index.

By the way, I have a general question related to this -r parameter:
Did you guys tried to omit this parameter (or as I did, accidentally forgot to put it)? I know the manual says "There is no default, and this parameter is required for paired end runs.", but it does work quite well if you forget it!
I tested with and without the option, on a same dataset and the difference was very marginal: 1 junction was not found without the parameter (out of >100,000) and 2000 alignments were missing (out of >700,000).
I have no idea what the default value is. The accepted_hits.bam file is well formatted and the reads are properly paired.
Similarly, I always wanted to test a bunch of values and see how sensible this parameter is. My assumption is "not very much", but I may be wrong. Did anybody tried?

Please let me know if I am the only one to get to work without the -r parameter.

[glados]

Hi Nicolas.

TopHat doesn't have much need for the mean inner distance nowadays, and will even less in the future, according to Cole Trapnell. If you don't specify -r with a value it will default to 50 as it says in the release notes from an earlier version, they just haven't updated the manual yet.

[glados]

I looked again at the Picard MarkDuplicates output and I think I may have interpreted the percent_duplication=0.250123 wrong. Perhaps they mean it is as high as 25%, not 0.25% as I initially thought. But when reading about what both samtools rmdup and Picardtools MarkDuplicates actually does: Marking reads as duplicates if they have the same 5' coordinates, then it doesn't seem strange that so many reads were marked as duplicates, considering that my inner distance is -50 and my reads are very overlapping so sometimes the reads will start on the same coordinate.

[swebb]

[QUOTE=Jon_Keats;68654]Couple of questions:

[lzlgboy]

I did some experiments for the same sample using Tophat 1.3.3. It seems the -r parameter of Tophat will have influence on cufflinks analysis. When I use a wrong -r parameter, the Tophat output.bam file change not much( you can see from IGV). But when it comes to cufflinks, some genes that have a FPKM value(read coverage in IGV) now have a zero FPKM. So I guess that, cufflinks will need the proper paired information to estimate gene's FPKM.
I am wondering if anyone observe that.

[figo1019]

setting up the -r parameter in tophat

Hi Nicolas.

TopHat doesn't have much need for the mean inner distance nowadays, and will even less in the future, according to Cole Trapnell. If you don't specify -r with a value it will default to 50 as it says in the release notes from an earlier version, they just haven't updated the manual yet.

Hi Glados

Reading the thread and your comments. I have a query about setting up the parameter in -r. I also tried to get the information of insert size from the sequencing lab and they said me it to be 180 bp for 96 bp illumina paired end reads. So what can be the ideal set up for -r for me will it be (180- (2x94))=-8. Do I need to have the -8 (aaprox -10) as - r or Another way could be like i keep it default.

Regards

[glados]

Dear figo1019.

Make sure the lab meant 180 as insert size and not inner distance. If I were you, I would try with both the default and -8 as inner distance. See if you get any differences in tophat output. You can view the assembly in a genome browser, like IGV, and see if the reads on average overlap approximately 8 bp. If you are certain it is -8, I would use it in tophat.

[hollyyang]

How to estimate the fragment length for single-end RNAseq reads

Hi Nicolas,

I have been stuck in estimating the fragment length for single-end RNAseq reads.

I applied your script (http://seqanswers.com/forums/showthread.php?t=16472) to my bowtie2 results but got no results.
my command:
samtools view -S -F 0x4 Ery_rep1.ip.sam | awk '{if ($9 >0) {sum+=$9;sumsq+=$9*$9;N+=1}} END {print "mean = " sum/N " SD=" sqrt(sumsq/N - (sum/N)**2)}'

output:
[samopen] SAM header is present: 22 sequences.
awk: (FILENAME=- FNR=18471674) fatal: division by zero attempted

My sam file looks like below, it seems that the inferred fragment lengths are to be zero. Is it because the my reads are single-end reads?

42A08AAXX:6:098:00152:01628/1 16 chr1 3011033 42 51M * 0 0 TCACCTGTTCTTCTCACTGTTGTGGCCTGAGTCAGAAC
AACTAGAGTCCTC ############################9.(<.6BA/@8>([email protected].= AS:i:-2 XN:i:0 XM:i:1 XO:i:0 XG:i:0 NM:i:1 MD:Z:19G31 -
YT:Z:UU
42A08AAXX:6:097:00856:00542/1 16 chr1 3011742 42 51M * 0 0 AGCTCTGTGTTCTGCTTGAGCTGACTCTCTAGACAGCT
ATGTGGATATTTC >;A:B>>B@?ABB@BB<@B>BBBBBBBABBBBBCBBBBBCBBBBCBCBBAB AS:i:0 XN:i:0 XM:i:0 XO:i:0 XG:i:0 NM:i:0 MD:Z:51 YT:Z:U
U


Thanks,
Holly