Some times people are worried about getting enough data from samples that contain starting low amounts of RNA (and/or get carried away) and do extra amplification steps. That will result in high duplicates in data. Do you know if this is applicable for your data?
what is the difference between duplicate pairs optical and non-optical?
What sequencer is your data from? Here is a long thread on biostars that has a useful image at top that shows the classification of duplicates: Duplicates on Illumina .
You can identify duplicates in an alignment free way using clumpify.sh from BBMap suite (Introducing Clumpify: Create 30% Smaller, Faster Gzipped Fastq Files. And remove duplicates. ) or after alignments using picard markduplicates.
Is the "high" expected duplicate levels referring to optical or non-optical or both types?
In general one expects there to be duplicates for RNAseq data since multiple copies of RNA's for a certain number of genes will be present in your sample. Chances of getting PCR dups is high if the samples are over amplified during lib prep. There are specific recommendations on loading to lower chances of getting optical duplicates (which most sequencing providers should be following by now).
Unless you have unique molecular indexes (UMIs) that were incorporated prior to amplification, there would be no easy way to detect PCR duplicates for certain. There are packages like https://pmc.ncbi.nlm.nih.gov/articles/PMC5073875/ that could be used.
