Why do some index sequences have an incorrect base?

[spawn]

Hello everyone,

I prepared a custom library using dual index TruSeq adapters for a Miseq run with a sample plexity of 3. This is the first time I performed library prep and I provided the information about indexes at the time of sample submission. For example, sample 1 has D501+D701 indexes, sample 2 has D502+D702 and so on. Today I received the data and it looks like it was already demultiplexed before they sent it to me as there were two separate files for each sample as Read1 and Read2.

In some reads, I see that a single base is incorrect. For this particular sample, the indexes should be GGAATCTC + GGCTCTGA. They appear as expected in the first read below but in the next one, the first "G" in GGAATCTC is replaced with a "T". While the majority of reads have the correct index sequences without any errors, could someone comment why some of them are off by a base? Is this normal and does this affect index assignment and demultiplexing?

Thank you.

@M00677:1216:000000000-K2G3W:1:1102:15780:1709 2:N:0:GGAATCTC+GGCTCTGA
GAGATTCCTGGGCGTCCTTAGTGCTCCGCTTAGAGTACTTCGTTTTACAGC
+
11>1>FFD3111A1AAAEGB1B1AFFC0AEE0111121DA1/0A//D1211

@M00677:1216:000000000-K2G3W:1:1102:13989:1710 2:N:0:TGAATCTC+GGCTCTGA
TAGATTCCGTTCATCTACTAGTGCTCCGCTTAGCTTACTGTTCTTTTTCCT
+
1111>DD31BCC1F3G3AB31B1BFDA0EEE011121BD2DB2DFFG1011

[luc]

You only had three samples on the run? Thus, the complexity of the index reads will be low and the quality of the index reads will be impacted due to extremely high fluorescence intensities in the lowest complexity parts. These result are expected.
By default the Illumina demultiplexing programs will tolerate one mismatch. There are no problems to be expected due to this.

You certainly have lost some reads due to demultiplexing problems (more than one mismatch). Potentially you could recover some of these by shortening the index information used for the demultiplexing. 16 nt of information is much more than needed to distinguish three samples.