Utilizing amplicon sequencing and qPCR results as an input to downstream pypgx analysis

[arnonl]

Hello good people,
I wonder what is needed to make a vcf from amplicon sequencing in combination with qPCR to serve as a legal input for pypgx analysis. I guess I cannot use one of the ready-to-use commands like the run-ngs-pipeline, but is there any combination of commands that can handle this?
Also, if applicable, what is a reasonable spacing between amplicon locations so the imputation has a reasonable quality?

Thank you!

[sbslee]

Hi @arnonl,

Thank you for your questions.

1. Am I correct in understanding that you have both amplicon sequencing data and qPCR data for the same sample(s)? PyPGx natively supports targeted sequencing data as input, though its effectiveness largely depends on the amplicon design. See this discussion for more details.
   BTW, do you have FASTQ and/or BAM file(s) for your sequencing data? If so, you can technically proceed using the run-ngs-pipeline command. Just make sure to specify --platform Targeted. You can read more about this command in the documentation here.

2. I also assume you're planning to use the qPCR data for CNV detection. Currently, PyPGx doesn't support qPCR data as a native input. However, if you can convert your qPCR data into CNV calls, you can supply those directly to PyPGx.
   If you refer to this figure in the NGS pipeline documentation, you'll notice a step where PyPGx outputs an archive file with the semantic type SampleTable[CNVCalls].
   You might be wondering how to map qPCR data to CNV calls—this issue is discussed in more detail here: Integration CNV/SV with PyPGx pipeline #56.

3. Regarding the imputation question, there are no straightforward answers, as imputation accuracy depends on several factors, with variant density being only one of them. If you can share more details about your amplicon design, I may be able to provide more specific guidance.

[arnonl]

Thank you very much Seung-been!

Here are my follow-up comments and questions:

1. Thank you. Yes, I have BAM files and I am aware of the '--platform Targeted' option. However, I wasn't sure how a targeted sequencing, like amplicon sequencing, without normalization and low coverage homogenity (like in capture sequencing, I suppose) gets the SV correctly. Thus I was wondering about getting a complementary support from qPCR.
2. Thank you for the references.
3. I would like to know more about the way to assure correct haplotype phasing. In the ClinPharmSeq study, for most cases you sequenced the whole exom and also the custom cases are mostly the whole gene. In this case you have enough "densed" data to achieve relaible phasing. However, do you have any experiance in downsampling the vcf file and the markers density and explore the effect on the correct phasing? Can one get similar quality while having a more spaced sequencing?

Thank you for your input.

[sbslee]

@arnonl,

Regarding your question about phasing, I haven’t tried downsampling a VCF file to assess how marker density affects phasing accuracy. However, as I mentioned in my previous reply, phasing accuracy depends on many factors, and marker density is just one of them. For example, allele frequency is another major factor. Even if two variants are far apart, if they frequently co-occur, the impact of their distance on phasing accuracy may be minimal.

Additionally, the choice of reference haplotype panel plays an important role. It's also worth noting that while phasing accuracy is important, the primary goal of PyPGx is to accurately call star alleles—not necessarily to phase every marker perfectly. This distinction matters because some markers are more critical than others, and phasing is especially important for those high-impact variants.

I hope this helps!

[arnonl]

True, very good points, thank you for highlighting these issues!