A computational algorithm and software tool for fast and accurate detection of gene fusion by long-read transcriptome sequencing
Long-read RNA-Seq techniques can generate reads that encompass a large proportion or the entire mRNA or cDNA molecules, so they are expected to address inherited limitations of short-read RNA-Seq techniques that generate only 50-150bp reads. However, there is a general lack of software tools for gene fusion detection from long-read RNA-seq data, which takes into account of the higher error rates and the possible alignment errors for long-read data. Here, we proposed a fast computational tool, LongGF, to efficiently detect candidate gene fusion from long-read RNA-seq data, including cDNA sequencing data and direct mRNA sequencing data.
A bam file of long-read transcriptome sequencing data sorted by name and a GTF file for the definition of genes. Please note that you bam must be sorted by NAME but NOT by POSITION.
You need to have C++ (GCC > 4.8.5) and HTSlib to compile and run the program.
There are several steps before you can run the program.
A single installation command of conda install -c bioconda longgf should work for you. If not, please post them on GitHub and try the installation below.
git clone https://github.com/WGLab/LongGFcd LongGF/bing++ -g -O3 -std=c++11 -I ./include -L ./lib -Wl,--enable-new-dtags,-rpath,"\$ORIGIN"/lib -lhts -o LongGF _com_fun_.c _gtf_struct_.c get_gfFrombam.c -Wl,--no-as-needed
Then, you will have LongGF/bin/LongGF to run.
You can run ./LongGF to get help documents about the parameters belw:
Usage: ./LongGF <input_bam> <input_gtf> <min-overlap-len> <bin_size> <min-map-len> [pseudogene:0(default)/1] [Secondary_alignment:0(default)] [min_sup_read:2(default)]
where
<input_bam>: A bam file. Compulsory.
<input_gtf>: A GTF file. Compulsory.
<min-overlap-len>: The minimum length of an alignment record overlap with a gene. Compulsory. Recommend: 100.
<bin_size>: The bin size during discretization. Compulsory. Recommended: 30, 50 or 100.
<min-map-len>: A minimum length of an alignment record against the reference genome. Compulsory. Recommend: 100, 200 or 300.
[pseudogene:0(default)/1]: Optional. Default=0: Not use pseudogene from the GTF file.
[Secondary_alignment:0(default)]: Optional. Default=0: not use secondary alignment.
[min_sup_read:2(default)]: Optional. Default=2.
The default output is in the standard output, and we recommend you using > LongGF.run.on.XXX.date.log after the running command so that the output will be directed to the log file. After that
grep "SumGF" LongGF.run.on.XXX.date.log
will give you a list of detected gene fusions. You can find more detail in the log file.
If you have any questions/issues/bugs, please post them on GitHub. They would also be helpful to other users.
Qian Liu, Yu Hu, Andres Stucky, Li Fang, Jiang F. Zhong, Kai Wang. LongGF: computational algorithm and software tool for fast and accurate detection of gene fusion by long-read transcriptome sequencing. BMC Genomics 21:793 (2020).