Good index sequences for unique dual indexing?

[jwfoley]

After the index-hopping scandal (see Illumina's white paper) it sounds like the field is moving in the direction of unique dual indexing, i.e. dual indexing where every index combination is unique, so that no two libraries have the same P7 index and no two libraries have thes same P5 index. The means that instead of 12 versions of the P7 adapter and 8 versions of the P5 adapter, you'd need 96 versions of P7 and 96 versions of P5. This is an expensive investment for homebrew library protocols.

Does anyone know of experimentally validated index sequences for 96-plexity?
It would be nice to keep the 2 x 8 nt index-read lengths so we don't have to steal cycles from the actual target sequence. I see Bioo has a set of 96 barcodes that are 8 nt long, so I guess you could just reuse those same 96 on each end. But 68 out of 96 have an A or C in the first position, and a whopping 90 out of 96 have a A/C in the last position. That seems like it would be a serious problem since A/C are read by the same-color laser and Illumina says you must have color balance in your barcode combinations, even on the older machines that do four imaging cycles.

[kmcarr]

I have heard unconfirmed reports that both Kapa and Bioo are planning on offering sets of 96 dual-unique indexes.

[MU Core]

I've been told that Illumina will soon (~1-2 months) be releasing a kit with 96 dual-unique indexes.

[jlei_face]

As you have probably already seen, Illumina will be working with IDT to create these UDI's. I know that IDT already has 96 unique indices, 8 nt long, as I have used them in the past. Instead of waiting for Illumina to release their kit it might be worth contacting IDT directly...

[apredeus]

Along the same lines, and rather than multiplying threads - are there any comprehensive reviews of all the indexing strategies available for Illumina?

And what would be a most reliable indexing strategy for single-end sequencing on Hiseq3000/Hiseq4000? Some long custom indexes, maybe?

I would appreciate any input or pointers to discussions on other sites or here.

[nucacidhunter]

I have not seen one. Generally it includes:
1- Indexes included in adapter (i5 and i7) which is Illumina’s standard method
2- Inline barcode where one can include a barcode sequence at the start of one or both of reads. Barcode can be fixed or variable length for different samples to increase sequence diversity if required.
3- Combinatorial indexing which is using inline barcode and standard index for samples

To minimise read miss-assignment due to index hopping one can dual index libraries with unique i5 and i7 sequences. Longer index will be inefficient in prevention or identification of index hopping.

[kmcarr]

Along the same lines, and rather than multiplying threads - are there any comprehensive reviews of all the indexing strategies available for Illumina?

And what would be a most reliable indexing strategy for single-end sequencing on Hiseq3000/Hiseq4000? Some long custom indexes, maybe?

I would appreciate any input or pointers to discussions on other sites or here.

You can use dual indexes (or dual unique indexes) even with single end reads. Here is a link to Illumina's index sequencing overview which describes all possibilities of indexing, sequencing and instruments.

[apredeus]

You can use dual indexes (or dual unique indexes) even with single end reads. Here is a link to Illumina's index sequencing overview which describes all possibilities of indexing, sequencing and instruments.

Thank you very much! This is extremely useful.

[Innovelty]

Still a little confused

I'm a bit confused about how to accomplish the dual-indexing using the same index on each end. My workflow ligates an adapter, and then uses indexing primers in PCR to get flow-cell adapters and indices onto the molecules. When I'm single indexing, this involves a "universal forward primer" and a unique reverse primer with the index. I can't just use the reverse primer without the universal forward primer, can I? I won't get indices on both ends, will I?

This seems awfully elementary, I know, but it's hard to keep track of exactly what protocol people are talking about doing. I 100% understand using two unique index primers for your library, out of, say, Illumina's TruSeq dual-indexing kit. But I don't understand how people are suggesting using the same index on both ends. You've got to be using custom oligos, right, with a forward and reverse primer that contain the same index region?

Are some sequencing cores now stocking indexing primers like this that customers can purchase? Because I sure as heck cannot afford 96 pairs of custom Illumina primers.

[luc]

You will require barcoded primers for both the P5 and and the P7 end of the library molecules. The "universal forward primer" needs to be substituted with a P5 index primer.
Yep, you will require custom oligos.

I'm a bit confused about how to accomplish the dual-indexing using the same index on each end. My workflow ligates an adapter, and then uses indexing primers in PCR to get flow-cell adapters and indices onto the molecules. When I'm single indexing, this involves a "universal forward primer" and a unique reverse primer with the index. I can't just use the reverse primer without the universal forward primer, can I? I won't get indices on both ends, will I?

This seems awfully elementary, I know, but it's hard to keep track of exactly what protocol people are talking about doing. I 100% understand using two unique index primers for your library, out of, say, Illumina's TruSeq dual-indexing kit. But I don't understand how people are suggesting using the same index on both ends. You've got to be using custom oligos, right, with a forward and reverse primer that contain the same index region?

Are some sequencing cores now stocking indexing primers like this that customers can purchase? Because I sure as heck cannot afford 96 pairs of custom Illumina primers.

[systembio]

Has anyone used IDT unique dual indexes on various Illumina sequencing platforms? Do UDIs behave differently on Miseq (v3 chemistry) vs HiseqX? Does a Miseq chemistry has a role to play in index representation?

[Innovelty]

Thanks for bumping this thread up again, systembio.

From what I know about the MiSeq chemistry (and it was the first sequencing platform I used) the unique dual indices shouldn't be a *problem* for MiSeq. Libraries made with UDI will sequence just fine on a MiSeq. They're unecessary, because MiSeq does not have the index-hopping problem that the HiSeq does, because the MiSeq's cluster-formation process is different. But definitely doing a little MiSeq nano run of your libraries to make sure they're what you wanted would be fine, prior to a HiSeq run. (If I'm wrong about this, though, I am sure someone will correct me!)

To answer my own question from a few months ago, the barcodes must be unique at both ends, but they don't have to be "the same index" at each end. Just each p5 index can only be used once, and each p7 index can only be used once, so each library gets a unique PAIR of indices. (Since I was struggling with it awhile back, I think I might as well clarify it for anyone else.) We ended up sequencing at Berkeley's genomics core, and they sell plates of UDI. Very pleased with them.

[nucacidhunter]

Has anyone used IDT unique dual indexes on various Illumina sequencing platforms? Do UDIs behave differently on Miseq (v3 chemistry) vs HiseqX? Does a Miseq chemistry has a role to play in index representation?

They work fine on NovaSeq and HS4000 and should be fine on MiSeq too as Innovelty has mentioned.

I wonder if you have used adapters directly from IDT or Illumina branded plates from Illumina. It will be interesting and useful if you could describe the issue that you have seen.

[systembio]

Thanks Innovelty and Nucacidhunter for your replies!
The problem i see here is that the libraries containing UDIs when an aliquot is run on Miseq show different index representation than the one run on HiseqX!!
i.e UDI pair poorly performing on Miseq is actually the best one on HiseqX! Why so?

[Innovelty]

I think it would be helpful to us to know where you got your indices from, as nucacidhunter mentioned. I don't know if it's possible, but perhaps the indices from Illumina are selected because they work very well on the HiSeq. Since the way that they cluster is different, I *could* see it being possible that certain sequences on the HiSeq perform better than on the MiSeq. But I really can't imagine why that would be so, unless certain parts of the sequence upstream of the flow-cell adapter are better for the recombinase enzyme that the HiSeq uses?

It could also be a function of the libraries themselves, and NOT the indices. Were your libraries all of uniform size distribution? I've observed that the MiSeq can be very sensitive to the presence of long (>800bp) inserts -- those molecules simply won't cluster as well as the shorter inserts. If the poor-on-MiSeq library had larger inserts than the other libraries, that could explain why it didn't form as many clusters on the MiSeq flowcell, but was able to form clusters well on the HiSeq.

I feel your frustration. Clustering is some kind of mystical voodoo to those of us who don't work at Genomics cores and don't have tons of experience with the way different libraries cluster on different platforms. I'm dealing with this myself right now -- wide spread of representation over 50+ libraries, only *partly* explained by an aggressive Pippin prep.