Hi! I am running

magmap

with

-profile docker

and I have my own genome files

.fna

and

.gff

in the

genomeinfo.csv

. I am getting an error with

NFCORE_MAGMAP:MAGMAP:PROKKAGFF2TSV

where the

ec_number

columnn is missing. The

.gff

file was downloaded from NCBI and I joined it. Any idea on what could be wrong?

Gff files are different between different annotation tools. I think the easiest for you would be to ignore the error since the output from this process is not strictly necessary but a "nice to have". This can be done by supplying a config file with the following content:

process {
    withName: PROKKAGFF2TSV {
        errorStrategy = 'ignore'
    }
}

The config file can be provided in different ways to Nextflow. Easiest, IMO, is to create a file called

nextflow.config

in the directory you're executing the pipeline from.

Thank you Daniel, with the

nextflow.config

the pipeline was successful 🙂

What are the proportions mapped to contigs and features respectively? There's a file that contains this info in the

results/summary_table

directory. And, did you run with defaults so you get rRNA and other non-mRNA features also quantified?

Yes I did run with defaults settings so I have the CDS anf other non-mRNA features. The

magmap.overall_stats.tsv.gz

look quite good with a high number of

idxs_n_mapped

(see picture)

Your mapping rates for CDSs also look very good; very close to the

idxs_n_mapped

. When you say above that you have a lot of

unassigned: No Features

where do you see that? Is it ORFs (CDSs) that are identified but not given a functional name? If so, this is part of the annotation of the genome and can possibly be helped by reannotating the genome with e.g. Bakta. I can help with that and resuming the pipeline with this as input. Or is it something in the MultiQC? Then, have a look at the other tables in

results/summary_tables

to see if they convey a different message.

The high numbers of

unassigned: No Features

come from the featureCounts part of the MultiQC report. That´s why I was a bit surprised when I saw the mapping rates for CDSs. When I look in the

magmap.prokka-annotations.tsv.gz

I have 648 out of 1560 orfs without gene name or cog. And 600 of them are CDSs. The strain used in the experiement already have a fully annotated genome (https://www.ncbi.nlm.nih.gov/datasets/genome/GCF_000016065.1/) so reannotating it would not make a lot of sense, isn't it?

I see. Then it's probably not the CDS run of featureCounts but one of the other, much lower. So, nothing to worry about.

Good to hear ! When I look in the

*_cds_from_genomic.fna.gz

file on NCBI, I have 7 genes of interest (N-cycle) that are annotated. However, only 2 were detected in my datase, and I was hoping to have more. In the

genomeinfoc.csv

I used the

*genomic.fna.gz

file, do you think trying a reannotation/specific tools could be beneficial?

You used the correct file to map to, i.e. the one with contigs. After mapping to contigs featureCount uses the gff file to understand which mappings to the contigs are within the boundaries of genes. I think all points to only two genes being sufficiently highly expressed to detect. Absence is of course problematic however. What makes you think these genes should be expressed? One alternative is to search for them directly in the reads. There are different approaches but I think Diamond in

blastp

mode would be easiest. You'd just align all clean reads to your genes of interest and see what you get. Alternatively, you could check the missing genes' coordinates and see if there's something that made featureCount miss them even if they're in the mapping data. That would require looking into the bam files.

My supervisor did a rather quick preliminary exploration of the data and found one more gene, thus I was expecting at least 3 genes. However I have found this "missing" gene with a slightly different name (

nrfA_2

instead of

nrfA

). This gene has 2 copies in the genome, I guess that explains the name difference. Now I suspect that it could be the case of other genes of interest