When aligning small RNAs (~20-24 nts in length) against the version 1 Norway Spruce genome (ftp://plantgenie.org/Data/ConGenIE/Picea_abies/v1.0/FASTA/GenomeAssemblies/Pabies1.0-genome.fa.gz .. over 10 million scaffolds, over 12Gb in length) :: observed that use of the --best option will produce alignments reported as perfect, but which are in fact not even related to the query sequence at all. Tested a variety of settings with a test fasta read (specifically, ATTAATGAACCCCCTCTAGGCCCC), and found the error was specific to the use of the --best option. Not sure how extensive the bug is (tedious to cross-check each alignment).
Probably tied to the .ebwtl (large) indices, as this is a bad bug and I've never seen it with smaller genomes (.ebwt indices).
bowtie-build was run on all default settings.
Testing on Mac OSX with pre-compiled binaries (see below)
bowtie version 1.1.2
64-bit
Built on anitra.peabody.jhu.edu
Tue Jun 23 12:36:27 EDT 2015
Compiler: Thread model: posix
Options: -O3 -m64 -stdlib=libstdc++ -DPOPCNT_CAPABILITY
Sizeof {int, long, long long, void*, size_t, off_t}: {4, 8, 8, 8, 8, 8}
bowtie-build version 1.1.2
64-bit
Built on anitra.peabody.jhu.edu
Tue Jun 23 12:36:15 EDT 2015
Compiler: Thread model: posix
Options: -O3 -m64 -stdlib=libstdc++ -DPOPCNT_CAPABILITY
Sizeof {int, long, long long, void*, size_t, off_t}: {4, 8, 8, 8, 8, 8}
btw, thanks for all the great work on bowtie !