[ptran]

You're right! I foolishly believed the documentation which starts with the statment:



It even says that if you don't provide a sample sheet it tries to read one from <input_dir/SampleSheet.csv>, with no mention that you can go without one all together!

I'd tried using a blank sample sheet (with no sample or project names) and that failed, but you can indeed not specify a sample sheet at all.

The v1.8 User guide gives the following scenarios

Bcl Conversion/Demultiplexing Examples
Bcl conversion and demultiplexing support four scenarios:

1)Multiplexed samples present, with sample sheet.
Reads are placed within the directory structure specified by the sample sheet, based
on the index and lane information. Reads for which the index sequence was
ambiguous will be placed in a project directory called Undetermined_indices,
unless the sample sheet specifies a specific sample and project for reads without
index in that lane.

2)Multiplexed and non-multiplexed samples present, with sample sheet.
Reads are placed within the directory structure specified by the sample sheet, based
on the index and lane information. Reads containing ambiguous or no barcodes
will be placed in a project directory called Undetermined_indices, unless the sample
sheet specifies a specific sample and project for reads without index in that lane.

3)No multiplexed samples present, with sample sheet.
Reads are placed within the directory structure directed by the sample sheet, based
on the lane information.

4)No multiplexed samples present, without sample sheet.
Reads are placed in a project directory named after the flow cell, and sample
directories based on the lane number.

[skruglyak]

I'm looking into this as well, and it seems that there are a few annoyances with the new version of CASAVA.

I suspect the reason for removing ANALYSIS sequence option is that it's not supposed to be needed since the default output format from base calling is now fastq, so you'd just not run alignment for that lane. However this leads to some headaches:

1. On some (possibly all?) platforms the output for a single lane is split into multiple fastq files which means you've got to merge these back together for most downstream analysis outside of CASAVA
   
2. All sequences are present in the fastq files, even ones which have failed QC. There is a flag in the header which says which sequences should be filtered, but no downstream applications understand this so you'll have to manually remove these entries before doing any further analysis.

Whilst Illumina have tried to move towards more standardised file formats (using Sanger offset fastq files and BAM for alignments), the new version of CASAVA is actually going to make our lives harder since we'll need to introduce a manual step to merge and filter the fastq files to get back to the single output file of good quality we had before.

If I'm reading the manual wrong with this (we haven't actually done a run under 1.8 yet), then I'd be very happy to hear it. How are other people coping with the changes in the new version?

Thanks for this feedback. Your concerns will be addressed in the minor October release. The default behavior will be to omit the reads that do not pass QC. For more detail, please see this post:

http://seqanswers.com/forums/showthread.php?p=52126#post52126

Additionally, you will have the option to produce 1 FASTQ file per sample.

thanks,

Semyon

[simonandrews]

Thanks for implementing this change. It's encouraging to see a company respond to users feedback.

[tonymell]

Using CASAVA 1.8.1. According to Illumina tech support, you can set --fastq-cluster-count up to 999,999,999.