search

nghiavtr / FuSeq

GNU General Public License v3.0
32 stars 12 forks source link

Split Read Processing Error #6

Closed mflevine closed 5 years ago

mflevine commented 5 years ago

Hi,

While testing FuSeq on a few more samples, I ran into an issue with one of my samples. I know that this sample has 4 confident fusions but I get an error after the split read processing says that there are 0 remaining reads. I have attached the stdout and stderr. Please let me know if I should send the FeqDir. FUSEQ_255244_command_log.log FUSEQ_255244_command_err.log

Best, Max

mflevine commented 5 years ago

It looks like this sample is short reads. I will explore if this because of the kmer size.

nghiavtr commented 5 years ago

Hi Max,

From FUSEQ_255244_command_log.log, I can see your data have 50bp read long, so it should use with kmer length of 21 to capture the split reads.

Best, Nghia

mflevine commented 5 years ago

Kmer length of 21 worked great! Are SR.passed and MR.passed supposed to always add up to supportRead? For many fusions I am seeing 0 in either SR.passed or MR.passed with supportRead being higher than the non-zero value. Please see attached. Thanks! FUSEQ_255244_fusions.FuSeq.txt

nghiavtr commented 5 years ago

The supportRead is the sum of all SR and MR reads including also SR and MR not passed very strict statistical tests and filters, thus supportRead>=SM.passed + MR.passed.

Either SR.passed or MR.passed with zero is simply because none of the reads passed our tests/filters (might be false negatives due to too strict conditions), or sometimes the read type has no clear evidence. However, the method found high confidence in the other read type.

Best, Nghia

mflevine commented 5 years ago

Thank you for the clarification!