.dict file created by picard and by samtools

[HeidiLee]

There are 2 ways of generating .dict file for human genome.

java -jar picard.jar CreateSequenceDictionary REFERENCE=reference.fa OUTPUT=reference.dict
samtools faidx ref.fasta

I found that the 2 files generated in the 2 different command have different file size.

I need .dict file for GATK and Picard. Which one is correct one?

Thank you very much.

[GenoMax]

Are you sure samtools faidx is creating a .dict file? It only creates an .fai file as I recollect.

[dvdleest]

There are actually 2 ways of generating .dict files.

Either

java -jar picard.jar CreateSequenceDictionary REFERENCE=reference.fa OUTPUT=reference.dict

or

samtools dict reference.fa -o reference.dict

Both commands are supposed to produce the exact same .dict file. However, due to a solved error in an older version of samtools dict, there can be subtle differences. In some contigs a few positions of the sequence were lost if they are not within the usual character set [ACTG], resulting in different sequence lengths and a different md5sum.

So to answer your question with respect to "which one is [the] correct one?"
Picard functions correctly, so either use picard or the updated version of samtools for making the fasta sequence dictionary.