Two Miseq runs - different quality

[tomlodz]

Hi everyone,

We try to do NGS on ancientDNA according to the protocol of this publication:
https://www.ncbi.nlm.nih.gov/pubmed/20516186

Library is prepared with custom adapters, single-indexed. The adapters used here resemble the most for the TruSeq LT kit. Also, custom primers are used for sequencing run.

We took two sequencing attempts on the V2 Nano 500 cycles Kit for MiSeq device. We did paired end reads 2x250. In both attempts, PhiX was added to the library.

In first run cluster density was far too low (100 mm2). This was probably due to the inhibition of the blunt-ending reaction in the first step of library preparation. What resulted in a low concentration of endogenous DNA. Beside this on my newbie eye overall run parameters are not perfect but acceptable.

In second run after real-time quantification concentration of the library was 5-6 x greater than previous. My calculations were correct because cluster density was now 500 mm2. Unfortunately, in this case the rest parameters of the run were horrible. Especially the first 10 cycles were very poor quality. This resulted in a lack of demultiplexing of the library because quality of Read 2 was far too low for analysis. I also notice that intensity was almost 10 x lower than the previous run but I don’t know reason why ?

Both runs was prepared similar. Also good too know that ancientDNA sequencing is associated with the presence of unknown lengths of the DNA inserts in the library.

Can you please provide some advice? What cause such difference in quality of reads for this two RUNS. Most of the parameters are included in the attached pdf file. Thanks a lot!

[GenoMax]

Were your indexes diverse (in terms of sequence)? What was the concentration of phiX added to this run?

[tomlodz]

In both cases 30 ul of 12,5 pM PhiX was added to the library.

In theory indexes were chosen that the read should be undisturbed. Below index sequences used in runs:
TCGCAGG
CTCTGCA
CCTAGGT
GGATCAA
GCAAGAT
ATGGAGA
CTCGATG
GCTCGAA
ACCAACT

In first attempt read of the indexes was quite OK.

[nettybetty]

Originally Posted by tomlodz

Hi everyone,

We try to do NGS on ancientDNA according to the protocol of this publication:
https://www.ncbi.nlm.nih.gov/pubmed/20516186

Can you please provide some advice? What cause such difference in quality of reads for this two RUNS. Most of the parameters are included in the attached pdf file. Thanks a lot!

I think a lot of the run difference can be explained by as you state: issues with quantification. You've included the MiSeq run stats, but I think there's a more important question - what did the library QC looks like (BioAnalyser/Gel)?Ancient DNA is massively variable in terms of quality and quantity. Were you able to visualise the DNA in extracts to determine mean size length prior to library prep? Often there is high molecular weight DNA in aDNA samples as well. A large variation in size of the library will impact qPCR and clustering. You should be able to easily see an ancient DNA library on a BioAnalyser chip post-PCR.

Endogenous DNA in truly ancient samples is always very low unless you enrich for it. The first Neanderthal genome sequence runs were around 90% microbial.

As a general rule - read 2 (second seq read) also is nearly always rubbish in ancient DNA for me on a HiSeq. I maybe get 20-30 bases before it turns to rubbish. I do know of a few possible reasons for this but won't go into detail to save text.

What is the concentration of adapters are you using? If there's not much DNA then the advised adapter concentration should be dropped to prevent dimers and artefact. Are you sure your aDNA is still double stranded? Meyer et al found a lot of single stranded DNA in their really old material and actually designed a new protocol to be able to sequence this. If your aDNA is really damaged it won't work very with the t4 ligation.

This protocol you are using is 7 years old and was designed for modern material (and derived from Meyer's et al. work on the 454). Are you changing anything e.g. this protocol shears the samples which is not advisable for most ancient DNA> A lot of this is no longer standard, even for modern material. Meyer and others have updated protocols to work much better for ancient DNA. You can also create ancient DNA libraries that don't require custom sequencing primer thus immediately removing a huge component of variability. And on that note - was your sequencing primer HPLC purified?

A 2x250bp run will result in most ancient DNA sequencing reads being full of adapter. Again - look at the library distribution in QC. Adapter sequence will mess up quality due to being low diversity etc. You're also paying for sequencing you're just going to throw out. On a Hiseq people will wear this when it's only a lane or two, but a MiSeq enables you to choose read lengths and customise a lot better and cut down work on the bioinformatics end.

[tomlodz]

Thank you for your answer.

Originally Posted by nettybetty

I think a lot of the run difference can be explained by as you state: issues with quantification. You've included the MiSeq run stats, but I think there's a more important question - what did the library QC looks like (BioAnalyser/Gel)?Ancient DNA is massively variable in terms of quality and quantity. Were you able to visualise the DNA in extracts to determine mean size length prior to library prep? Often there is high molecular weight DNA in aDNA samples as well. A large variation in size of the library will impact qPCR and clustering. You should be able to easily see an ancient DNA library on a BioAnalyser chip post-PCR.

I realize that the quality of aDNA and its length may be a problem here. I visualized the library after indexing on PAGE gel electrophoresis. The maximal length of inserts seemed to be about 500-600 bps depending on the age of the material. I'm talking here about the ones that were visible on the gel. There was a strong smear between 140 - 600 bp. Certainly there were longer fragments of modern DNA contamination present. Is there a way to get rid of them without losing a large part of the library. The second question is whether the larger DNA fragments have a very strong impact on the sequencing reaction itself.

Originally Posted by nettybetty

Endogenous DNA in truly ancient samples is always very low unless you enrich for it. The first Neanderthal genome sequence runs were around 90% microbial.

As a general rule - read 2 (second seq read) also is nearly always rubbish in ancient DNA for me on a HiSeq. I maybe get 20-30 bases before it turns to rubbish. I do know of a few possible reasons for this but won't go into detail to save text.

My pre-enrichment library had about 1000x more copies than after this operation. The obtained readings was mapped in most cases to the mitochondrial genome rather than the nuclear genome. After this I conclude the enrichment reaction was OK.

Originally Posted by nettybetty

What is the concentration of adapters are you using? If there's not much DNA then the advised adapter concentration should be dropped to prevent dimers and artefact. Are you sure your aDNA is still double stranded? Meyer et al found a lot of single stranded DNA in their really old material and actually designed a new protocol to be able to sequence this. If your aDNA is really damaged it won't work very with the t4 ligation.

I adjust the adapter concentration to the aDNA concentration. In this experiment I used material from several historical periods whose degree of preservation should be sufficient to create a library. Oldest sample was about 6000 years old.

Originally Posted by nettybetty

This protocol you are using is 7 years old and was designed for modern material (and derived from Meyer's et al. work on the 454). Are you changing anything e.g. this protocol shears the samples which is not advisable for most ancient DNA> A lot of this is no longer standard, even for modern material. Meyer and others have updated protocols to work much better for ancient DNA. You can also create ancient DNA libraries that don't require custom sequencing primer thus immediately removing a huge component of variability. And on that note - was your sequencing primer HPLC purified?

I searched the literature for what methods other teams use, and most use the ones mentioned here with minor modifications. There are more sophisticated methods to the aDNA material whose preservation state is very bad. Material used here I was able to investigate with the standard Sanger method and PCR amplification (186 bp). All used oliginucleotides were HPLC purified.

Originally Posted by nettybetty

A 2x250bp run will result in most ancient DNA sequencing reads being full of adapter. Again - look at the library distribution in QC. Adapter sequence will mess up quality due to being low diversity etc. You're also paying for sequencing you're just going to throw out. On a Hiseq people will wear this when it's only a lane or two, but a MiSeq enables you to choose read lengths and customise a lot better and cut down work on the bioinformatics end.

In the first run I actually had a lot of adapter dimmers because of failed blunting reaction of aDNA. In the second run adapter-dimers were present but not in such large numbers. After the reaction I removed the adapters present in the readings.

P.S.
In the second run the intensity was very low. What factors have the greatest influence on this parameter ?

[tomlodz]

I'm still fighting with sequencing. I noticed a rather strange dependence which I can not explain for now.

In this thread I wrote that sequencing of my second library resulted in tragic parameters. Then I thought it could be caused by faulty reagents or equipment. Part of this library left me so I tried to set it again but its concentration was much smaller. The sequencing parameters of the second less concentrated library were very good (in attachment parameters of this two runs) .

So I prepared two more libraries and unfortunately the script repeats itself. If the library has a higher concentration, especially fluorescence intensity is tragic and the passing filter is very low (about 8-10%). The readings for Phix are OK in these reactions. I excise the last library from agarose gel to get rid of long fragments which can disturb cluster generation but without any positive result.

From what I have read, this low intensity is usually related to the problem of the Read1 primer failure. But how to explain that if there is less library then everything works well. I really have no idea what is wrong with this.

Maybe someone had a similar problem.

[Markiyan]

In this case first try having a look at the raw thumbnail/image data, and make a movie from the run's tiles (using imagemagic, one frame per cycle).

This can help a lot in understanding what actually is going on.
Usually the read quality is better with lower cluster densities on the Miseq/Hiseq2500/NextSeq instruments.

Also it is a good idea to capture the raw images when working with ancient DNA.
(So they can be reprocessed later).

In case of MiSeq it would mean doing some storage mods: attaching an external DAS array and mounting is as drive D: Ideally you want to use esata type one set in RAID 10 like 8-16TB of useable capacity and make sure to use 64KB clusters for NTFS volume or use IFSD driver with ext4 and large allocate.

[tomlodz]

OK, When I have access to the Miseq, I will stick the thumbnail images into a movie and post it.

Unfortunately, the modifications of data storage in my case are not an option because it is not our device but thank you for showing this possibility.

[tomlodz]

I have stick the thumbnail images into a movie from this two runs.

Low concentration of library and good overall quality:
https://youtu.be/0utJYdPv_8o

High concentration of the same library and horrible quality:
https://youtu.be/vv1mzw0tvjM

[GenoMax]

Has Illumina tech support looked at this run/your MiSeq? I am wondering if there is some kind of an hardware problem with your sequencer.

[thermophile]

Both of those look very underclustered. No where near 500k. One tiny thing that can ruin your cluster density is old NaOH. When was the last time you've made 2N from 10N stock? You're diluting the 2N fresh each time right?

[tomlodz]

GenoMax
Illumin's support did not verify these results. It is not our device and it is only granted to us.

thermophile
Exactly both reactions are underclustered only what I am thinking about is such drastic differences in the quality of readings of both reactions of the same library at different concentrations.
Each time a new dilution of NaOH is made from 10M stock.

[GenoMax]

You should have Illumina tech support look at this data. It looks like you have access to the entire set of raw data so you should be able to supply files that tech support is going to ask you to provide. Send an email to techsupport at illumina.com. Describe this problem to get started.