Strange interactive quality plots and DADA2 filtered out a large number of sequences

[Geng_Xinxing]

After importing my data, I got some strange interactive quality plots. I checked the forum and saw that this might be normal, so I went ahead and used DADA2 for denoising. However, I noticed that a large number of sequences were filtered out. I've used DADA2 before and got good results, and I usually only use DADA2 in Qiime2. How should I proceed with my analysis from here? Thanks a lot!

[colinvwood (Colin Wood)]

Hello @Geng_Xinxing,

The quality plots look like that likely because your sequencing technology used the newer "binned" quality score approach--this is nothing to be concerned about. As to why so many reads were filtered, can you attach the dada2-stats qvz and your demux qzv?

[Geng_Xinxing]

Thanks so much for your helpful answers! I've uploaded
dada2_denoising-stats.qzv (1.2 MB)
paired-end-demux.qzv (311.9 KB)
files. Do you have any suggestions for the analysis I'm working on afterward? Should I try some data preprocessing before continuing with DADA2 to get better results? Or maybe use a different method for denoising?

[colinvwood (Colin Wood)]

Hello @Geng_Xinxing,

The problem is that you provided truncation parameters that are longer than many (most) of your reads. If you look closely at the help text for the --p-trunc-len-* parameters you will see:

"Reads that are shorter than this value will be discarded."

Look at the "Demultiplexed sequence length summary" table in the demux visualizer and then adjust the truncation parameters. (You can probably drop these parameters altogether since there is no quality drop off in your reads.)

[Geng_Xinxing]

Oh, thank you so much! You really reminded me. I had set --p-trunc-len to 125, and it ended up filtering out a lot of sequences. Now I changed it to 100, and everything's working fine. Thanks again!