Calculating percent error expressed in geoduck transcriptome

[ocattau]

Lets say my data looks like this table after the count data has been aligned with the blast data. In order to calculate % error expression, my first reaction would be to calculate % NA in the Identifier column, but sometimes there is count data without an identifier.

My second guess would be to filter the tpm so only values > 0 are counted, then calculate the % that has any value at all, but would some values be NA and some 0?

I don't think it makes any sense to calculate % error expressed using the Target_ID column since most of them have values but sometimes there are identifiers that don't have a corresponding Target_ID.

If I use the R function complete.cases, it removed any row with NAs which would delete all rows 1-3 in this example, and in my mind if there is count data but no identifiers, wouldn't that mean the gene exists but we haven't identified it yet? Any help with this data wrangling would be appreciated.

Target_ID	Identifiers	tpm
NA	BJC123	NA
Trinity-X12	NA	20
Trinity-Y99	NA	0
Trinity-Z33	ABC321	3000
Trinity-Z34	ABC321	2000

[sr320]

Can you explain what you mean by "% error expression"?

[ocattau]

So I should just ignore any TargetID without an identifier? (for example I would delete the first three rows in this sample data set)

[sr320]

You keep every Target ID in the table I provided

[ocattau]

In that case 95% of the data set is NA, is that normal?

[sr320]

You want to know the expression pattern of every Target ID, including those that are not expressed at all.

[sr320]

so you really have a line like line 1? no target ID?

[sr320]

I think I see the problem

You should be using left_join (or something similar) where you retain only annotation information that matches data table with expression

[ocattau]

[sr320]

we do - but your line 1 has no count data.

[sr320]

left join would not remove line 2 and 3.

[ocattau]

https://github.com/ocattau/code/blob/main/characterize_larval_transciptome/inspecting%20kallisto%20output%20SR%20merged%20kallisto.Rmd

[ocattau]

Okay I figured it out, I did need to add a left.join at line 42 of the code I posted above, thanks!

[sr320]

Put a link to the GitHub repo where you are working.. that would be easiest way to understand and troubleshoot

[sr320]

what does you slurm header look like?

[ocattau]

[ocattau]

I think I need to wait till Sam's job is done because of memory issues

[sr320]

change mem to 100G

[sr320]

Full explanation - we have two nodes
https://robertslab.github.io/resources/Computing-Hardware/#mox-hyak

Sam is using the 500M, but you do not require that. When you set mem to <120G your job will start on 2nd node.