NGM - Next-generation EMS mutation mapping - FAQ

Frequently Asked Questions:

Preprocessing

Does NGM work in species other than Arabidopsis?

Theoretically yes, it should work for any species with a complete genome and a usable mapping population. It has been used to successfully map dozens of Arabidopsis mutations and does allow for data generated from several different species genomes.

Can I use a mapping line of an ecotype other than Landsberg?

Definitely, though its yet to be done in practice and Columbia as the mutant line is needed. However, the principal for NGM is exactly the same. NGM will use the natural variation existing between the two lines to identify the Columbia specific non-recombinant region as a SNP desert.

Can I use a mutagenesis line of an ecotype other than Columbia?

Yes, NGM now allows for a number of different Arabidopsis ecotypes as provided by the 1000 genomes project.

The genomic background I'm mapping against is not available in your drop down, can I still perform NGM?

Yes, you just need to send us your genomic fasta file, one sequence per chromosome, along with an associated GFF3 gene model file where the chromosome numbers in the GFF correspond to the chromosome numbers in the fasta annotation line of the genome. These should all be a single integer with no extra characters (eg. "Chr"). You will also need to provide the source URL for both files so that other users can make use of the same data.

Can I map a dominant mutation?

Again, theoretically yes. You would need to take a larger population of F2s to F3, saving F2 tissue from each individual. F2 tissue would then be taken for corresponding F3 populations no longer segregating for wild-type.

Can I map a non-EMS induced mutation

While NGM doesn't support indel polymorphism annotation or the use of indels as de novo generated markers, the bulk segregation stage could potentially still be used to rough map the mutation. Using an indel calling program against this region in your BAM file and intersecting with relevant GFF data could be used to generate a short-list of candidate genes.

Should I back-cross or use F3 tissue?

We advise against it. In cases where this has been done, the SNP signal corresponding to EMS derived discordance appears diluted, peak identification and annotation issues crop up and mutation detection is more difficult.

Can I use an EMS mutagenized introgression line or mutant background and find enhancers/suppressors?

Yes, although linkage with the background mutation on the same chromosome may render the fine mapping step ineffectual. You should try to approximate the non-recombinant region using the SNP desert in the SNP frequency histogram if you encounter problems.

How do I create a SNP file?

Generate a SNP file by mapping your NGS data against a reference genome. NGM supports MAQ,SAM and VCF format data. Using samtools to create a snp file from pileup data (Samtools 0.16 or earlier) can be accomplished with:

	"samtools pileup -vcf reference.fasta alignment.bam > out.pileup"
	

where your reference matches one of the available references in the drop down and alignment.bam is created using any BAM compliant mapping method.

As of 2013 it is recommended to use a VCF/BCF file as generated with the below command and converted to an NGM file using the Perl script provided on the NGM start page.

		"samtools mpileup -E -ugf reference.fasta alignment.bam | bcftools view -bvgN - > output.bcf"
		

Using MAQ, preprocessing requires a pileup file and a cns2snp file generated using:

	maq pileup chr.bfa align.map > maq.pileup
	maq cns2snp in.cns > out.snp
	

Should I filter my SNP file before preprocessing?

We advise strongly against it unless you are working with a NGM file generated from BCF data. With samtools v0.16 or earlier, NGM seems to work best with the default MAQ/Samtools output and the conservative filters provided on the start page. Moreover, aggressive filtering may remove the actual causal mutation such that your results are sunk before you even begin. NGM allows the user to remove filters at the final annotation stage in order to liberally consider all possible SNPs in the event that a strong signal isn't obtained.

I'm having trouble with the applet.

Yes, on some systems, issues with the browser, Java version or browser preferences are unfortunately common. Our best advice is to make sure cookies, java and java script are turned on and you click "Allow" at the Security dialog. Other than that, make sure your java is up to date, and you try different browsers and operating systems if available. If this fails, Perl scripts are available for download preprocessing your file to "emap" format locally.

As the applet is specific to pre-BCF data, it is recommended to not use the applet but use the Perl script for processing BCF data on the home page. The applet will be discontinued in the near future.

What is this preprocessing thing?

NGM uses a chastity statistic pegged to the reference genome called discordant chastity to classify SNP allelism and discordance with the reference genome. This information is added to the SNP file and extraneous data removed before compressing the results for upload.

What is an emap file?

Emap is the prototype file extension and remains vestigial. ".ngm" is the extension generated by the BCF2NGM Perl script for processing BCF/VCF data.

Bulk segregant analysis

I don't see a SNP desert in any of the histograms.

In the Arabidopsis x Landberg mappings to date, the very reproducible natural variation pattern can be seen in the examples provided on the NGM start page. Visible SNP deserts as they appear in Arabidopsis (Col) x Arabidopsis (Ler) are usually very apparent. You may try using more stringent filters to help improve the visibility of the depletion.

I still don't see a SNP desert in my histograms.

First, make sure your experimental design is accurate with respect to the reference genome mapped and that chosen for Annotation. If your convinced your seeing a real natural variation pattern, but perhaps noisy, then run NGM against each chromosome and note the max ratio obtained at the default k-value and smallest kernel size. The max ratio on the non-recombinant chromosome will frequently be orders of magnitude larger than the other chromosomes.

Chastity band ratios

Will my mutation be at the top of a peak?

Ideally yes, but NGM uses a ratio taken from two smoothed estimations of SNP frequency to localize the mutation. Thus, depending on the available markers,signal strength and parameters used this may not always be the case. Sometimes the causal mutation will be slightly to right or left of a peak. Also, using too fine a kernel may cause "peak shifts" away from the actual mutation.

Does the height of the peak reflect the probability?

No, any peaks are a result of taking the ratio of two estimations and have no statistical likelihood associated with them. False peaks and shifted over-fit peaks are possible, especially as you approach the telomere.

What effect do telomeres have on peak identification

Unfortunately, if your mutation is near a telomere, the use of peaks for localizing the causal mutation becomes considerably less accurate. This is because you don't get multiple recombination events on either side of the mutation but one-sided events that result in a drop-off in recombination as you approach the telomere. In this case, the entire telomere should be annotated based on the depletion seen in the frequency histogram.

Is every peak a potential mutation?

No, there should be one best peak that approximates your causal mutation.

I don't see any distinct peaks after running the chastity belt partitioning

Click on detailed view to see NGM run at multiple parameter settings.

I think I see a strong peak, what do I do?

Click on detailed view to see NGM run at multiple parameter settings to try and fine tune the peak.

Ok, I have a strong peak at the default settings.

Congrats, you likely nailed it. Click on the arrow beside the kernel size of choice.

I've clicked on detailed view and I get roughly the same peak across the parameter array.

An unambigous result is good, pick one and fine tune your annotation.

How do I use these parameter settings?

Generally, lower k-values will incorporate more marker information and are preferred over large k-values. Similarly, smaller kernel values will apply finer and finer estimations that may overfit the data such that larger kernel sizes are better. One should raise the k-value to generate better band separation or lower kernel values to improve refinement of existing peaks (see below).

I've got peaks at multiple places and multiple peaks within a single chromosome.

Begin at the lowest k value of k=5 in the left column. Proceed from the largest kernel size down to the smallest looking for a distinctive peak. If a peak "shifts" or appears suddenly at lower kernel sizes without having precedence at larger kernel sizes, than move to the next greater k value and repeat. Stop when you see a well defined peak possessing precedence at larger kernel sizes.

The progression from k to smaller kernel sizes doesn't really work.

See if a quorum exists at larger kernel sizes for peak position. Annotate SNPs within a large window in that region.

Ya .. I just don't see a strong peak.

Identify the non-recombinant region on the histogram and annotate the entire corresponding region in the fine mapping. Gradually move the red sliders inwards to obtain a short-list. Depending on the size of your mapping population, mutation position and number of recombination events, the non-recombinant region often takes on a parabolic shape with the causal mutation at the apex.

Whats with all the peaks and the tweaking, can't this fit into a probalistic framework?

Probably, but the supervised empirical approach of NGM is analogous to traditional fine mapping by design and works very well as is. The application of chastity belts to more complicated mapping problems is ongoing research and may incorporate a more probabilistic approach though.

Annotation

I just see a swirly ball after adjusting my annotation filters.

The adjustment of annotation filters can take several minutes. Usually this is a matter of being patient but see below.

I don't see my annotation details. Why is that?

This should hopefully be fixed but is most often the result of the first column of your ".emap" file containing strings like "Chr". It should be a single integer corresponding to the chromosome number.

When I lower my discordance cutoff, I see mutations with no base change. Why is this?

This is a consequence of Samtools' use of ambiguity codes at polymorphic loci with lower chastity scores. As a result, NGM ignores the ambiguity code and will consequently filter the SNP as a transversion. Pulling the major allele information from such positions is on the TODO list. If you have dirty data and are using a lowered discordance cutoff than you should uncheck the transversion filter box and consider all C->Y and G->R changes as (potentially non-synonymous) transitions.

Why are transversions being filtered out?

EMS mutagenesis deaminates cytosines to tyrosines, causing transition mutations at a rate over 99% in Arabidopsis and very high in other species. Thus the causal mutation is highly likely to be a transition.

Why are there filter settings in the annotation region?

If you have poor mapping quality or coverage, you can remove or loosen the filter settings to see if potential candidates exist with more lenient filtration.

I get multiple potential candidate mutations. Now what?

You should prioritize multiple candidates for testing in the lab. You should consider 1) the proximity of the mutation to the peak and/or the apex of the non-recombinant depletion, 2) the BLOSUM score significance for residue change, 3) the annotation of the gene in the context of your analysis, 4) the position of the mutation within the gene identified and 5) whether the mutated position is a conserved position in other ecotypes/strains of the same species.

Are these other mutations I see real?

Likely yes, most are probably the EMS induced changes that have been carried along due to linkage with the causal mutation. They may also reflect polymorphic differences between your lab line and the sequenced reference line.

What is the BLOSUM score about?

The BLOSUM score is pulled from the BLOSUM100 amino acid substition matrix. Roughly speaking, the larger the number, the more dramatic the substition and likelihood for a disruptive effect on the peptide.