library insert size: too small

[capricy]

Hello,

I got some illumina data which obviously has very small insert size as many of the read pairs have >95% overlap. I went back to illumina library protocol and feel that it generally suggests a library insert size longer than the total length of the paired-end read, that is, read pair does not overlap.

What is the optimal library insert size? Should I move forward for data analysis?

Thanks a lot!

[GenoMax]

What kind of an experiment is this? It may still be fine to use that data after appropriate trimming. You could optionally merge.

[capricy]

It is metagenomic data from fecal samples. I did trim the samples with trimmomatic.