Which is the best platform for sequencing target genes in a large number of samples?

[DCD]

Hi everyone!

Our lab wants to start a high-throughput sequencing project but we are undecided about which platform is the most suitable for our purposes (we are new to NGS field).

Our prior aim is sequencing about 400 human samples (200 cases and 200 controls) in a bunch of approximately 15 selected genes related to our studied disease. We have previously performed GWAS and/or Ichip assays on these samples. The point here is that we want to sequence the entire gene (introns and exons) and about 250kb upstream and downstream of the gene (~9Mb total for 15 genes) with 40x depth of coverage. Moreover, we need a high accuracy and the possibility of multiplexing the highest number of samples in each run. Low length reads aren't a problem, as well as high run times or long procedures.

So, which would be the most suitable platform for us? Illumina, SOLiD, Ion...?

Any advice or guidance would be appreciated!

[Brian Bushnell]

I'd suggest Illumina, as it has the lowest error rates, and can multiplex 96 samples per lane. The amount of sequence you need (144 Gbp) could be handled by a few HiSeq lanes, running at 2x150bp; so you could probably do everything you need with 4 or 5 lanes. Or a few MiSeq runs which would give you even longer reads, up to 2x300bp. Longer reads, at the same error rate, are more accurate in mapping.

[DCD]

I'd suggest Illumina, as it has the lowest error rates, and can multiplex 96 samples per lane. The amount of sequence you need (144 Gbp) could be handled by a few HiSeq lanes, running at 2x150bp; so you could probably do everything you need with 4 or 5 lanes. Or a few MiSeq runs which would give you even longer reads, up to 2x300bp. Longer reads, at the same error rate, are more accurate in mapping.

Thanks for your reply!

[nucacidhunter]

First step in the experiment would be to obtain those 9 Mb sequences including genes of interest and flanking region from your samples. The best option is to use custom sequence capture kits which are currently offered by few commercial vendors. Kits are designed to capture target region for sequencing on various NGS platforms and currently the best option is using kits for capture and sequencing on Illumina systems. After preparing libraries from samples they can be multiplexed pre-capture (up to 16 samples/ reaction) and captured sequences can be pooled further post-capture for sequencing. Sequencing needs would be dependent on the required depth for downstream analysis. On target capture efficiency is usually above 60%.

[pmiguel]

First step in the experiment would be to obtain those 9 Mb sequences including genes of interest and flanking region from your samples. The best option is to use custom sequence capture kits which are currently offered by few commercial vendors. Kits are designed to capture target region for sequencing on various NGS platforms and currently the best option is using kits for capture and sequencing on Illumina systems. After preparing libraries from samples they can be multiplexed pre-capture (up to 16 samples/ reaction) and captured sequences can be pooled further post-capture for sequencing. Sequencing needs would be dependent on the required depth for downstream analysis. On target capture efficiency is usually above 60%.

Would that be the on target capture efficiency for exomes? Seems like nearly any several hundred kb chunk of a mammalian genome will be replete with repetitive DNA. Feed oligos for capturing those segments and you would end up skyrocketing your off target hits.

You could repeatmask the sequence prior to generating the capture oligo sequences, but then you will not have contiguous sequence.

I don't know a solution to this. Maybe long term, capture oligos to the single copy segments of interest could pull out ~50kb segments of genomic DNA that could be sequenced with something like a MinION?

Problems there as well as single molecule sequencing currently results in high raw errors rates.
--
Phillip

[nucacidhunter]

Would that be the on target capture efficiency for exomes? Seems like nearly any several hundred kb chunk of a mammalian genome will be replete with repetitive DNA. Feed oligos for capturing those segments and you would end up skyrocketing your off target hits.

In sales flyers it is usually exome, but I have had 55%-90% of captured fragments being on target using custom capture kits from non-model organism. On target reads defined as at least being 80% homologous to target region in a stretch of 60 bases. Using repeat region as target capture probes most likely will increase the on target portion of captured fragments for two reasons: 1) any captured read from repeat region wether from target area or some other region would be on target as defined above. 2) Non-captured fragments from repeat region (ssDNA can bind to magnetic streptavidin beads) also will be classified as on target. As large portion of genome is repetitive, it is highly likely that non-probe bound fragments which are eluted would be repeat sequences. Not that I am advocating capturing repeat region.

I don't know a solution to this. Maybe long term, capture oligos to the single copy segments of interest could pull out ~50kb segments of genomic DNA that could be sequenced with something like a MinION?

Pulling out long fragments with probes designed to non-repeat region still will capture other regions containing the same repeat as well by daisy chain mechanism. Perhaps blocking repeat region with another set of non-biotinylated probes or Cot DNA before addition of capture probes can overcome this problem. Although by sequencing whole fragment those non-specific ones can be filtered out bioinformatically. Other issue would be preparing dsDNA from captured long ssDNA for sequencing. Current platforms require dsDNA as input for library prep.

[pmiguel]

In sales flyers it is usually exome, but I have had 55%-90% of captured fragments being on target using custom capture kits from non-model organism. On target reads defined as at least being 80% homologous to target region in a stretch of 60 bases. Using repeat region as target capture probes most likely will increase the on target portion of captured fragments for two reasons: 1) any captured read from repeat region wether from target area or some other region would be on target as defined above. 2) Non-captured fragments from repeat region (ssDNA can bind to magnetic streptavidin beads) also will be classified as on target. As large portion of genome is repetitive, it is highly likely that non-probe bound fragments which are eluted would be repeat sequences. Not that I am advocating capturing repeat region.



Pulling out long fragments with probes designed to non-repeat region still will capture other regions containing the same repeat as well by daisy chain mechanism. Perhaps blocking repeat region with another set of non-biotinylated probes or Cot DNA before addition of capture probes can overcome this problem. Although by sequencing whole fragment those non-specific ones can be filtered out bioinformatically. Other issue would be preparing dsDNA from captured long ssDNA for sequencing. Current platforms require dsDNA as input for library prep.

The low Cot blocking could work, except it would be hard to avoid sequencing the low Cot blocking probes. Unless the low Cot DNA could be converted to RNA. Low Cot cRNA. Random prime reverse trancription of low Cot DNA. Denature and treat with a DNAse. Clean-up and recover your Low Cot cRNA.

--
Phillip