GATK RNA-Seq pathway Split 'n' Trim- problems with .bai and .fai index compatibility

[gwilymh]

I am processing aligned RNA-Seq data to get phased haplotypes of novel genes in sea stars. I am working through the analysis pipeline that used the Genome Analysis Toolkit suite of software, and is detailed on the Broad Institute website:
https://www.broadinstitute.org/gatk/...ces?bpm=RNAseq

and

https://www.broadinstitute.org/gatk/...c?name=methods

I am up to the Split 'n' Trim stage of the pipeline, where the SplitNCigarReads tool is used to refine the sequence alignment. I have hit some problems with my index files. The order of the scaffolds and contigs for .fai (fasta index file) and .bai (index for bam file; bam file has been sorted, read-groups have been added and duplicated marked) files is different.

e.g. the .fai file:
Scaffold1 154793 11 70 71
Scaffold2 383464 157027 70 71
Scaffold3 336159 545981 70 71

eg. the .bai file (samtools idxstats filename.bam):
Scaffold1 154793 0 0
Scaffold10 244803 0 0
Scaffold100 200181 0 0

I therefore got the following error from GATK:
##### ERROR MESSAGE: Input files reads and reference have incompatible contigs: Relative ordering of overlapping contigs differs, which is unsafe.
##### ERROR reads contigs = [Scaffold1, Scaffold10, Scaffold100, Scaffold1000...
##### ERROR reference contigs = [Scaffold1, Scaffold2, Scaffold3, Scaffold4…

I wrote a script to change the order of the items in the .fai file so they would be in the same order as the .bai file, but this produced another error:
##### ERROR MESSAGE: Couldn't read file filename.fa because Mismatch between sequence dictionary fasta index for filename.fa, sequence 'Scaffold2' != 'Scaffold10'.

Can anyone suggest a good work around for this problem? Do I need to change the order of entries in the fasta file reference sequence? Should I alter the .dict file (scaffolds/contigs are in the same order as they are in the original .fai file)?

[dpryan]

You can't just reorder the lines in the fai file, they need to be in the exact same order as the contigs in the fasta file.

You have two options:
(1) Reorder the fasta file. If there are a LOT of contigs then this could be a pain. The simplest method is to split the fasta file by chromosome and then reconcatenate things...in the proper order.
(2) Reorder the BAM file. You can use picard tool's ReorderSam command to do this. This will order things to match the fasta file.

I expect option (2) is easier in your case.