Basespace update

[thermophile]

Illumina replaced MiSeq reporter as the demultiplexor for miseq data on basespace with bcl2fastq a couple of weeks ago. Since then I've run into a number of instances where MSR and bcl2fastq are different which had meant many demultiplexing failures. To save people time I thought I'd start the list of issues I've hit.

1. MSR allowed "." in sample names and sample ID, bcl2fastq does not


2. MSR treats "N" as wildcard, bcl2fastq treats it as exact. I run a mix of dual 8bp and TruSeq lt single index 6bp. MSR had allowed me to just put NNNNNNNN as the i5 and NN at the end of the i7, this no longer works. you have to use the actual sequences (AT at the end of i7 and TCTTTCCC for i5)


3. bcl2fastq or basespace is much much slower at demultiplexing. It used to take <30min to rerun a sample sheet, it's taking >4hours now.


4. bcl2fastq doesn't allow you to set the indexing mismatch (at least tech support that I talked to didn't know how to globally set). It tries to allow 1 mismatch and only drops to exact match if the hamming distance is <3

[kcchan]

In addition the BaseSpace apps are no longer free. You now must be subscribed to a Professional account to access the apps and also pay each time you run them.

[microgirl123]

Interesting about having to pay to use all of the apps. We're going to have some unhappy customers!

[thermophile]

Well that sucks, I just talked a few users into trying BaseSpace based on the NCBI_SRA app

[fanli]

Originally Posted by thermophile

3. bcl2fastq or basespace is much much slower at demultiplexing. It used to take <30min to rerun a sample sheet, it's taking >4hours now.

Perhaps this is related to the number of cores you allow bcl2fastq to use?

Originally Posted by thermophile

4. bcl2fastq doesn't allow you to set the indexing mismatch (at least tech support that I talked to didn't know how to globally set). It tries to allow 1 mismatch and only drops to exact match if the hamming distance is <3

from the bcl2fastq --help text:

  --barcode-mismatches arg (=1)
number of allowed mismatches per index
multiple entries, comma delimited entries, allowed; 
each entry is applied to the corresponding index;
last entry applies to all remaining indices

there is also this, which I have no idea what it does:

  --adapter-stringency arg (=0.9)                 adapter stringency

[kcchan]

Originally Posted by microgirl123

Interesting about having to pay to use all of the apps. We're going to have some unhappy customers!

I would imagine some of the third party developers aren't too happy either since their apps are now stuck behind a paywall.

[thermophile]

$5k to upgrade to professional which gives you the privilege of paying for apps

[thermophile]

Originally Posted by fanli

Perhaps this is related to the number of cores you allow bcl2fastq to use?


from the bcl2fastq --help text:

  --barcode-mismatches arg (=1)
number of allowed mismatches per index
multiple entries, comma delimited entries, allowed; 
each entry is applied to the corresponding index;
last entry applies to all remaining indices

there is also this, which I have no idea what it does:

  --adapter-stringency arg (=0.9)                 adapter stringency

Thanks! I'll have to see if i can change something in the sample sheet to set this

[GenoMax]

Originally Posted by thermophile

Thanks! I'll have to see if i can change something in the sample sheet to set this

Or switch to using bcl2fastq locally instead of BaseSpace

[thermophile]

Originally Posted by GenoMax

Or switch to using bcl2fastq locally instead of BaseSpace

I may have to do that, but that means I'll have to build a server for distributing the data to clients.

[GenoMax]

Originally Posted by thermophile

I may have to do that, but that means I'll have to build a server for distributing the data to clients.

If you are part of an academic institution then look into tapping common central compute resource. That way you won't need to become a sys admin in addition to other hats you wear (and not have to worry about security etc). If your users use that central compute resource then they would appreciate getting their data directly delivered to them.

[fanli]

Originally Posted by GenoMax

If you are part of an academic institution then look into tapping common central compute resource. That way you won't need to become a sys admin in addition to other hats you wear (and not have to worry about security etc). If your users use that central compute resource then they would appreciate getting their data directly delivered to them.

My experience has been that you always still need a bit of sysadmin experience to configure things exactly the way you like. For example, how do you add a new user/client for data access? Sometimes it's just easier to htpasswd it yourself

[elutheria]

the only charge for the apps is the cost for the compute on AWS unless it is a 3rd party app that costs to run. Also there are still Free accounts that come with some credits so you can trial Basespace. If your clients plan on using it a lot for analysis then they will need to upgrade otherwise they can still receive the data on a free account i think.

[Geneus]

Originally Posted by elutheria

the only charge for the apps is the cost for the compute on AWS unless it is a 3rd party app that costs to run. Also there are still Free accounts that come with some credits so you can trial Basespace. If your clients plan on using it a lot for analysis then they will need to upgrade otherwise they can still receive the data on a free account i think.

Or buy an Illumina sequencer and negotiate free use of BaseSpace for some time period as part of the deal.

[ScottC]

A couple of things:

- BaseSpace will still do basecalling for free from instrument runs.
- The newest version of bcl2fastq2 will now treat N bases properly (as 'wildcards' so-to-speak)

Also, surely this doesn't come as a surprise... as far as I'm aware, it was pretty well communicated a long time ago that it was going to become a pay-per-use service.

Cheers,

Scott.

[thermophile]

I'm not surprised that they moved to subscription. But an annual $5000 buy in to give someone the opportunity to pay to use apps is ridiculous.

[GenoMax]

Originally Posted by thermophile

I'm not surprised that they moved to subscription. But an annual $5000 buy in to give someone the opportunity to pay to use apps is ridiculous.

To be fair, that $5k (if that is what the cost of the mid-tier subscription is) appears to include two additional features.

Whether "Multi-user access" and 8hr of bioinformatics professional services support is worth $5K, is open to debate If you don't have any local IT infrastructure then that may be your sole option.

[kcchan]

Originally Posted by ScottC

- The newest version of bcl2fastq2 will now treat N bases properly (as 'wildcards' so-to-speak)

That issue still hasn't been fixed as of the the most recent 2.18 update.

[GenoMax]

Originally Posted by kcchan

That issue still hasn't been fixed as of the the most recent 2.18 update.

Software release note for 2.18 has this listed as "known issue"

“N” is incorrectly allowed as an index sequence character in the sample sheet. When used, this will cause a mismatch for any sequence character other than “N”.

[Jessica_L]

Originally Posted by kcchan

I would imagine some of the third party developers aren't too happy either since their apps are now stuck behind a paywall.

Does that apply to all apps or just the third party ones that we would normally be paying for? Right now, it looks like the usual basic analyses I'm used to running are still accessible from a basic account, but I think my account is still on a trial basis from the changeover.