search

AmpliconSuite / AmpliconSuite-pipeline

A quickstart tool for AmpliconArchitect. Performs all preliminary steps (alignment, CNV calling, seed interval detection) required prior to running AmpliconArchitect. Previously called PrepareAA.
Other
58 stars 28 forks source link

Tumor-only vs Normal-paired #63

Closed yaosichao0915 closed 1 month ago

yaosichao0915 commented 3 months ago

I noticed that AA can work with tumor-only WGS data since it was tested on the cell line data. I am just wondering what would be the benefits if giving with paired normal WGS as well.

jluebeck commented 3 months ago

Hi, this is a great question. While AA itself will not used the paired normal BAM, the focal amplification seed region detection stage performed by CNVKit will utilize the normal bam and this may help improve quality of CNV calls used for seeding - particularly in reducing false positive amplifications.

Jens

yaosichao0915 commented 3 months ago

Hi, this is a great question. While AA itself will not used the paired normal BAM, the focal amplification seed region detection stage performed by CNVKit will utilize the normal bam and this may help improve quality of CNV calls used for seeding - particularly in reducing false positive amplifications.

Jens

Thanks Jens. One Further question on this topic, since you mentioned false positive amplifications. Is there any tips to figure out the false positive? like low copy number or not on coding genes or through the amplicon picture provided?

jluebeck commented 3 months ago

If you see the same exact event in multiple unrelated samples, this can be an indicator of a false positive (similarity score reported by AmpliconClassifier may help establish this). Events with highly uneven CN number in low-complexity regions of the genome are definitely suspect amplicons - you may see something like this in the AA sashimi plots. We have otherwise put a large amount of work into building a database AA uses to filter problematic parts of the genome, so the false-positive rate should be low already.

newpest commented 3 months ago

Hi Jens, I have a small question. We found that the Amplicons obtained when using different Caller seeds as input will be very different. For example, when using ReadDepth and Cnvkit, there will be many different Amplicons with low average copy ratios. Should these Amplicons be ignored?

jluebeck commented 3 months ago

Yes, the caller used for seed identification can influence the results. ReadDepth is now deprecated and we recommend CNVkit for the seed identification stage. I would be happy to take a look and provide feedback on some of these discrepant results if you would like to share. I would imagine most of these are falling into the "linear" class.