Low demultiplex yield

Question

[Steven]

Hello everyone,

I have a problem during demultiplexing and I don't know where to ask since I am new to bioinformatics.

So, I need to construct QTL map. I have 3 DNA libraries, each library consists of 68 samples (I extracted DNA and constructed library myself), using ddGBS method, and sequencing was done with Illumina HiSeq. I used axe-demultiplex and AdapterRemoval to demultiplex my data and turned out the yield was pretty low (around 50% when mismatch = 1). Because my barcodes are not equal in length (some 5bp and some others 6 bp), I split the demultiplex process into 2 according to the barcode length and run with AdapterRemoval. Turned out the yield was around 85% when mismatch = 2.

#axe demultiplex script
/home/guest/gst24-21/anaconda3/envs/stevenyodas/bin/axe-demux 
    -m 1 
    -p ON 
    -b /home/guest/gst24-21/scratch/ca10gb/library1.txt 
    -f /home/guest/gst24-21/scratch/ca10gb/P1_223C3LLT4_R1.fastq.gz 
    -r /home/guest/gst24-21/scratch/ca10gb/P1_223C3LLT4_R2.fastq.gz 
    -F /home/guest/gst24-21/scratch/ca10gb/demux2lib/ 
    -R /home/guest/gst24-21/scratch/ca10gb/demux2lib/ 
    -t /home/guest/gst24-21/scratch/ca10gb/demux2lib/demuxlib1.csv

#AdapterRemoval script
AdapterRemoval --file1 /home/guest/gst24-21/scratch/ca10gb/P1_223C3LLT4_R1.fastq.gz 
               --file2 /home/guest/gst24-21/scratch/ca10gb/P1_223C3LLT4_R2.fastq.gz 
               --barcode-list /home/guest/gst24-21/scratch/ca10gb/lib15.txt 
               --barcode-mm 1 
               --basename lib15 
               --demultiplex-only

My questions:

1. Can i split demultiplex into 2 based on the length?
2. I acknowledge that mismatch = 2 means my demultiplex result is not as reliable as mismatch = 1. I want to know your opinion about setting mismatch = 2 when the barcode length itself is short.
3. Is it a dry or wet bench problem?

Thank you