aln2wig error with pacbio long-read BAM and PSL files

[MatteoSchiavinato]

Hi,

I'm implementing a long-read-based gene prediction pipeline making use of Augustus/3.2.3. I'm using aln2wig to generate a wig track from the pacbio alignment. The alignment was conducted with BLAT so it's in PSL format. However, I am getting this error:

*** Error in `aln2wig': double free or corruption (!prev): 0x0000000001d5d320 ***
======= Backtrace: =========
/lib64/libc.so.6(+0x81489)[0x2b579a917489]
../scripts/augustus_scripts/aln2wig/aln2wig[0x4015e1]
/lib64/libc.so.6(__libc_start_main+0xf5)[0x2b579a8b83d5]
../scripts/augustus_scripts/aln2wig/aln2wig[0x401736]
======= Memory map: ========
00400000-00402000 r-xp 00000000 00:33 10372488375622258018               aln2wig
00602000-00603000 r--p 00002000 00:33 10372488375622258018               aln2wig
00603000-00604000 rw-p 00003000 00:33 10372488375622258018               aln2wig
01d5d000-048cc000 rw-p 00000000 00:00 0                                  [heap]
2b579a672000-2b579a694000 r-xp 00000000 fd:00 3410242                    /usr/lib64/ld-2.17.so
2b579a694000-2b579a698000 rw-p 00000000 00:00 0 
2b579a6ac000-2b579a6af000 rw-p 00000000 00:00 0 
2b579a893000-2b579a894000 r--p 00021000 fd:00 3410242                    /usr/lib64/ld-2.17.so
2b579a894000-2b579a895000 rw-p 00022000 fd:00 3410242                    /usr/lib64/ld-2.17.so
2b579a895000-2b579a896000 rw-p 00000000 00:00 0 
2b579a896000-2b579aa58000 r-xp 00000000 fd:00 3411761                    /usr/lib64/libc-2.17.so
2b579aa58000-2b579ac58000 ---p 001c2000 fd:00 3411761                    /usr/lib64/libc-2.17.so
2b579ac58000-2b579ac5c000 r--p 001c2000 fd:00 3411761                    /usr/lib64/libc-2.17.so
2b579ac5c000-2b579ac5e000 rw-p 001c6000 fd:00 3411761                    /usr/lib64/libc-2.17.so
2b579ac5e000-2b579ac63000 rw-p 00000000 00:00 0 
2b579ac63000-2b579ac78000 r-xp 00000000 fd:00 3419817                    /usr/lib64/libgcc_s-4.8.5-20150702.so.1
2b579ac78000-2b579ae77000 ---p 00015000 fd:00 3419817                    /usr/lib64/libgcc_s-4.8.5-20150702.so.1
2b579ae77000-2b579ae78000 r--p 00014000 fd:00 3419817                    /usr/lib64/libgcc_s-4.8.5-20150702.so.1
2b579ae78000-2b579ae79000 rw-p 00015000 fd:00 3419817                    /usr/lib64/libgcc_s-4.8.5-20150702.so.1
2b579c000000-2b579c021000 rw-p 00000000 00:00 0 
2b579c021000-2b57a0000000 ---p 00000000 00:00 0 
7ffc60e23000-7ffc60f29000 rw-p 00000000 00:00 0                          [stack]
7ffc60ffa000-7ffc60ffc000 r-xp 00000000 00:00 0                          [vdso]
ffffffffff600000-ffffffffff601000 r-xp 00000000 00:00 0                  [vsyscall]
Aborted

This happens consistently, and only with pacbio alignments (with short illumina reads I always get the right output). This error does not come out if I use bam2wig from a bam alignment. Am I missing something?

[KatharinaHoff]

I recommend to not use bam2wig for this purpose. You are going to miss valuable information that long read transcript alignments can provide: the chaining of different transcript parts. Have a look at how we handle EST and cDNA data for gene prediction with AUGUSTUS. I recommend going that way.