Upgrading our index set

[pezz...@gmail.com]

Dear all,

We have been successfully generating Smart-3Seq libraries in 96 well format using our set of legacy i7 indexing primers. However, we are ready to upgrade to 384-well format and was hoping for updated guidance on what indexing/sequencing scheme to use. Any suggestions are greatly welcome:

1) We only have access to a NovaSeq6000, does anyone have experience using NovaSeq for the i5 vs i7 indexing scheme?

2) Has anyone, for simplicity/storage, used combinatorial indexing rather than UDIs for Smart-3Seq?

3) Do you have available any index set that would achieve this?

4) Any new recommendations on how to order from IDT (e.g. purification method, ultramers vs not, etc)

Separate update: Whereas initially we used huge amounts of PhiX (up to 40%!) to prevent errors from the GGG reads in cycles 6,7,8 under the legacy scheme (with 96 primer set), we have consistently shown that 5% phiX works well, and plan to test even lower.

Thank you very much!

Alejandro Pezzulo

[Joe Foley]

We only have access to a NovaSeq6000, does anyone have experience using NovaSeq for the i5 vs i7 indexing scheme?

I don't but I asked Illumina about it and they said they'd get back to me very soon, so stay tuned. I imagine i5 indexing will require the same workaround as on the MiSeq and NextSeq: just do the minimum number of cycles in Index Read 1 (i7) so the software doesn't complain (3 wasted cycles on MiSeq, 2 on NextSeq, waiting for answer about NovaSeq).

Also, I'm not experienced with the NovaSeq myself, but I gather the "Xp Workflow" allows you to run different library pools in each lane (2 lanes on SP, S1, S2; 4 lanes on S4). So keep in mind that you can run e.g. 192 or 384 libraries together using only 96 indexes, though on the larger flow cells that might still be more reads per library than you really need.

Has anyone, for simplicity/storage, used combinatorial indexing rather than UDIs for Smart-3Seq?

This could certainly save some money on the PCR primers, however the reason we stopped using i7 indexes altogether and switched to 96-plex i5-only indexing is because we seem to get a lot of poly(A) artifacts in the i7 index read and not in the i5. So you might lose some fraction of usable reads if you're depending even partly on the i7 index.

It may be confusing and risk human error (including bad choices of index combinations) if you manually combine a unique P5 and P7 primer in every sample at the time you do the experiment, rather than premix all possible combinations so it's just one tube of primer pair to one tube of sample, unless maybe you're doing a full plate and going down the rows and columns with a multichannel pipet. If you have to make 384 tubes of premixed primers anyway, then combinatorial indexing doesn't really save so much trouble, just cost.

Do you have available any index set that would achieve this?

IDT (xGen) and PerkinElmer (NEXTFLEX) already sell UDIs up to 1536-plex (2 x 10 nt), so if you can track down their designs I could put those into a list of designs for i5-only indexing. Or it's possible you could just buy their PCR primer sets if they're TruSeq-compatible. I'm not sure if anyone has 384-plex combinatorial dual indexes; you could of course choose 16 x 24 of the UDIs, but these indexes can be surprisingly fussy about cross-compatibility so that's risky and needs empirical validation.

Any new recommendations on how to order from IDT (e.g. purification method, ultramers vs not, etc)

On IDT's advice we've switched to Ultramer synthesis with no purification, and only one phosphorothioate at the end. The Ultramer synthesis is so efficient that it doesn't really need purification to get mostly full-length oligos, and for PCR primers it's actually not a big problem if some of them are slightly truncated anyway. So skipping the purification greatly reduces the cost with little downside. (Keep the RNase-free HPLC for the 1S and 2S primers, though; those need it.)

I see IDT has a smaller 200 pmol scale for Ultramer oligos ordered in plates (theoretically up to 40 libraries per index), so depending on the size of your project that might be economical. It seems to be the same price to order them "wet" (in TE) rather than lyophilized, which saves a lot of time on that scale, though I still remeasure every single tube on the Nanodrop because of a traumatic experience years ago.

Whereas initially we used huge amounts of PhiX (up to 40%!) to prevent errors from the GGG reads in cycles 6,7,8 under the legacy scheme (with 96 primer set), we have consistently shown that 5% phiX works well, and plan to test even lower.

Thanks very much for this information, which will definitely be helpful to other users. On other platforms (MiSeq, NextSeq) it works fine with no phiX at all, because it's only the GGG cycles that have low diversity. I've heard the NovaSeq may be more sensitive to losing focus in the dark GGG cycles, so it may still require a little bit of phiX. But even if you find that it doesn't, I don't recommend skipping phiX altogether, because it's a cheap insurance policy in case of instrument failure - tech support's first response is to blame the homemade libraries unless you can show them phiX didn't work either.

--
You received this message because you are subscribed to the Google Groups "Smart-3SEQ" group.
To unsubscribe from this group and stop receiving emails from it, send an email to smart-3seq+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/smart-3seq/ee4eff99-3015-4266-a15f-f9a437779ad4n%40googlegroups.com.

[Joe Foley]

I looked into this some more and found some options. The designs I got from IDT for their xGen UDI primers do appear equivalent to the PCR primers in the Smart-3SEQ protocol:

• #10008052 Plates 1-4 (384 index pairs), $1152
• #10008053 Plates 1-8 (768 index pairs), $2304
• #10008054 Plates 1-16 (1536 index pairs), $4608
  

These are 5 µL @ 10 µM each primer in the pair (50 pmol each), plus overage, which is enough for 10 reactions per index pair with protocol version 1.9. Thus the 384-pair kit is enough for 3840 libraries @ $0.30 each. They also sell much larger aliquots of the same plates.

I'm not planning to test these myself, but if you want to be the pioneer, maybe you can first verify that these primers are compatible with Smart-3SEQ by testing the smallest version, #1000597 (16 pairs, $65), before committing to a set of 384. It would be interesting to sequence both index reads and then see whether the results get better or worse when you ignore the i7.

For comparison, a 384-well plate of custom Ultramer primers @ 200 pmol (40 rxn each index, 15,360 total rxn) would cost about $4700 ($0.31/rxn), assuming no i7 index, but those syntheses scale up more gradually. So the choice may depend on convenience and scale.

To view this discussion on the web visit https://groups.google.com/d/msgid/smart-3seq/2a74d9bb-56b0-596a-91f8-0e61011ea944%40stanford.edu.

[Joe Foley]

I don't but I asked Illumina about it and they said they'd get back to me very soon, so stay tuned. I imagine i5 indexing will require the same workaround as on the MiSeq and NextSeq: just do the minimum number of cycles in Index Read 1 (i7) so the software doesn't complain (3 wasted cycles on MiSeq, 2 on NextSeq, waiting for answer about NovaSeq).

Finally got the answer on this: on the NovaSeq, waste 3 cycles in index read 1 (some software versions may not need that many but this is the safest recommendation). So if you're doing e.g. a "100 cycle" run, it would be

• 101 cycles read 1
• 3 cycles index read 1 (i7, ignore)
  
• 10 cycles index read 2 (if you're using 10 nt 384-plex i5 indexes)
• (no read 2)
  

Please let everyone know here if you try this.

To view this discussion on the web visit https://groups.google.com/d/msgid/smart-3seq/4a3a9faf-1909-5fb3-f979-76f209c49c18%40stanford.edu.

[pezz...@gmail.com]

Dear Joe

Thanks so much, this is really helpful. I'll go over with the rest of the group and decide whether we'll switch to i5 indexing, and make changes to the sequencing scheme. You brought good points, we may be better off sticking to sets of 96 and reuse in lanes, and switch to 384 only if we are constantly making those libraries; in fact, this may help our core avoiding waiting for libraries to load in the other lanes.

Thanks again

Alejandro

[Joe Foley]

in fact, this may help our core avoiding waiting for libraries to load in the other lanes.

On this point, just a general notice for everyone: the recommended sequencing configuration for Smart-3SEQ (single-end reads) is not very commonly used these days, since most other libraries come from protocols (especially genomic DNA resequencing) that make use of paired-end reads for one reason or another. So if you are using a larger machine with divisible flow cells like a NovaSeq or an antique HiSeq, you probably have to commit to ordering an entire flow cell for yourself. Check with your sequencing facility to be sure; they'll know what's common in their queue. As an alternative maybe you can just do unnecessary paired-end sequencing anyway and discard read 2, but you'll have to check the pricing to figure out whether this actually saves you money compared with reserving an entire smaller flow cell.

To view this discussion on the web visit https://groups.google.com/d/msgid/smart-3seq/4529e132-5e2c-4cb9-80db-b76cacfa3d51n%40googlegroups.com.