Adjusting the defaults in the cn.MOPs Package
Entering edit mode
voksdvoks • 0
Last seen 7.2 years ago

Hi all,

I'm curious to know what the default settings in the cn.MOPs package are, for example, normalization, thresholds and priorImpact? Can I possibly play around with them? How do I access them? Any assistance will be greatly appreciated.



cn.mops cnv cnvtools • 1.4k views
Entering edit mode
Last seen 20 months ago

Hello OAO,

These are the parameters that determine precision/recall/sensitivity/specificity of the method.

a) Normalization: Corrects for the samples' different coverages. This is completely analog to the normalization problem in RNA-Seq. With the default parameters the median read count per sample is used to estimate the coverage of a sample. Normalization is difficult if there are CNVs that affect a large proportion of the genome as it happens in certain cancer types. For detection of germline CNVs the default parameters should work well.

b) PriorImpact: cn.MOPS uses the prior assumption that a sample has copy number 2. Only if the data speaks very much against this assumption, the model explains the data with another copy number (e.g. CN1 or CN3). The higher you set priorImpact, the fewer copy numbers you will find. The default setting is 1, which should work well in many settings, you might consider changing the parameter on the logarithmic scale: 0.1,1,10,100, etc..

c) Thresholds: This is technically not a parameter of cn.MOPS. This is just for reporting the results. cn.MOPS returns a full segmentation of the genome in the slot "segmentation". Here both normal segments (CN2) and segments with copy number changes are reported. For each segment there is an average "sI/NI" (signed individual informative/non-informative call) reported. The farer the value is away from 0, the more likely there is a CNV. The parameters "upperThreshold" and "lowerThreshold" only report segments with an average sI/NI value above "upperThreshold" or below "lowerThreshold". Overall, increasing the absolute value of the threshold will lead to less reported CNVs, and vice versa.

Can I possibly play around with them? Yes, you can. You should also consider playing around with the window size ("WL" parameter of "getReadCounts.." and the parameter "minWidth"). If you have a trainings set, you should adjust these values to optimize performance on your training set.

How do I access them? All mentioned parameters, except "WL", are arguments of the function "cn.mops". You can just set them to the value you want to test.

I hope this helps - let me know if you need further assistance!





Entering edit mode

Thanks Gunter. This helps.




Login before adding your answer.

Traffic: 451 users visited in the last hour
Help About
Access RSS

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6