Is there a more elegant solution to the bwa-mem: paired reads have different names error?

Question

I'm currently trying to run bwa-mem on Influenza substrains using the following command:

~/bwa mem h5n1_1_cons.fa h5n1_1_read1.fq h5n1_1_read2.fq

h5n1_1_cons.fa is the consensus sequence for substrain h5n1_1, and the fq files are paired end reads 1 and 2. Running the program gives me this error:

[mem_sam_pe] paired reads have different names: "JQ714243-920-1", "JQ714243-920-2"

Visual inspection shows that the read names are identical until the last character. I have found a package (bbmap) with a utility, bbrename.sh, that renames the sequences, producing:

./bbrename.sh in=h5n1_1_read1.fq out=fixed prefix=h5n1_1

which yields the following output

-bash-4.2$ head fixh5n1_1_read1.fq
@h5n1_0
gaaaatgtctcaggagttgatgcatggtgtttaatctttggtttgcgctg
+
@4>C5?7FA?9@@@;1,-4B+E??4E9,DA>6CB)*2.)?7/.'65+75/
@h5n1_1
catcagttggattttgcctaaggatgtccaccatccttatcccaccaatt
+
7(9?)=C3AEF+@@F=C-E>EE57CD:;'?.*?7;>:3>2.;29..)=2'
@h5n1_2
catgcatttgttggaataatgttctcacaaatcaactgtattgaccgctg

This fixes the problem and allows BWA-MEM to work. However, header information is discarded. Is there another way to ensure the header names are the same while retaining all header information?

Answer 1

Assuming the read name lines only contain things like @JQ714243-920-1 and nothing else, then sed -i '1~4 s/-[12]$//' file.fastq (N.B., example updated a bit for safety, thanks to @terdon) will remove the -1 and -2 suffixes.

However, you should try to figure out why the read names were munged into that form to begin with (at least Illumina machines won't produce that, I don't know about others).