Combining fast.gz files

[ebolyen (Evan Bolyen)]

Hey @d.davidburmeister,

@d.davidburmeister

I am truly sorry to keep bugging you. I promise I tried to troubleshoot this.

It's what we're here for

@d.davidburmeister

Unrecognized file (CombinedRuns/RS76B_S158_L001_R1_001.fastq2.gz) for CasavaOneEightSingleLanePerSampleDirFmt

The issue is in your file extension: it says "fastq2.gz" instead of "fastq.gz". QIIME 2 isn't sure what the 2 means, and so it gives up assuming the difference has some semantic significance.

Just rename and you should be set!

[d.davidburmeister (David)]

Thank you - this worked!

Was able to confirm there was no difference between runs (alpha or beta).

So now I am at the fun part, but want to ensure I treat the data correctly. I used the group plugin colin mentioned (grouped around sample axis) to create a new FeatureTable.

My question is- I don’t need to make a new FeatureData/ representative sequences file to analyze diversity, correct? I can simply use the new grouped feature table, and a new metadata file that has all samples grouped?

Thanks SO much!!

David

[colinbrislawn (Colin J Brislawn)]

Hey David,

@d.davidburmeister

I don’t need to make a new FeatureData/ representative sequences file to analyze diversity, correct?

That's right! You can use the table with grouped replicates, and a metadata file that also groups the replicates together.

Well done. You are making good progress,
Colin

[d.davidburmeister (David)]

Thanks- could not be done without you guys!

I am getting more comfortable with this work flow. Just to confirm my understanding- if I wanted to analyze the beta diversity significance for a certain variable (for example, cancer lesion size) but had a set of samples that did not make sense for that variable (in example above, because they are normal, healthy controls), I could simply make a trimmed down metadata file that does not include those samples, but still use the same FeatureTable, yes?

Seriously, thanks to you all for all of your guidance!

David

[colinbrislawn (Colin J Brislawn)]

Hello David,

Yes, that’s right. It’s pretty common to make subsets of your data to test for various things. Now that you have your OTU table, you can make subsets of samples at will.

Depending on the stat test you are doing, you could also pass in a blocking variable instead of subsetting your table. Take a look at how this linear-mixed-effect-models example let’s them test for delivery,diet,sex in their fully blocked study design.
https://docs.qiime2.org/2018.6/tutorials/longitudinal/#linear-mixed-effect-models

I’m not an expert on the stats so I’ll let the qiime devs recommend more.

Keep in touch!
Colin

[Nicholas_Bokulich]

@d.davidburmeister

I wanted to analyze the beta diversity significance for a certain variable (for example, cancer lesion size) but had a set of samples that did not make sense for that variable (in example above, because they are normal, healthy controls), I could simply make a trimmed down metadata file that does not include those samples, but still use the same FeatureTable, yes?

Other way around, actually — use qiime feature-table filter-samples to subset your feature table, but you can use the same metadata file as long as it contains a superset of samples in the feature table.

I hope that helps!