Different results with the same quality process with DADA2

Hi :slight_smile:
I have processed three batches of samples into three tables and now I want to combine these tables. The tables are from the same samples taken in different times, I have different names in each time ( e.g. T1P1_T0, T1P1_T1 and T1P1_T3) . What I did was to merged all the files before import them into Qiime2 and then I processed them as I did with each batch before with the same parameters with DADA2. But the results aren´t the same when I did the process individually (each batch) and when I did the process with all the reads in a unique batch. What I found reading the forum was that I have the option to merged the features tables and the rep-seqs files but I had the doubt about it. If I do the three batches individually isn´t the same if I merged the reads and apply the same process inside DADA2?

For example, the sequence counts individually are:
T0: 409.438
T1: 433.124
T3: 436.366

but when I processed all in a unique batch the sum of the T0_T1_T3 times are: 1.289.019 and it shoud be 1.278.928 ¿How can pass 10.091 seqs with the same quality process?
I attached the files :

table-16S-single-end-tiempo-3-20qs-180ml-180t-10trim.qzv (461.9 KB)
table-16S-single-end-tiempo-1-20qs-180ml-180t-10trim.qzv (436.3 KB)
table-single-end-t0-20qs-180ml-180t-10trim-1.qzv (439.1 KB)
table-single-end-t0-t1-t3-20qs-180ml-180t-10trim.qzv (579.1 KB)


Were your three batches done in different runs? DADA2 estimates the error rate from the data, and the DADA2 paper does mention batch-to-batch variation in error parameters. If your batches were sequenced in separate runs, they could each have a slightly different set of error parameters. This would lead to different results when you combine all three runs and then complete DADA2. An alternative possibility is because when you combine all three batches you have more sequences, DADA2 is better able to identify rare variants. This discussion on the DADA2 forum may help you to understand what is going.


This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.