We are trying to develop a method to call somatic CNV from cfDNA (cancer patient) with control.
However, I can't think of any method that have enough sensitivity to do that. Even we can track the allele frequency in the original library with the barcoding, the AF is pretty low (probably less than 1%). Even the copy number of somatic CNV is high (say 8 copies), the copy number is still diluted to 8% increases (8 * 1%) and highly unlikely to be detected.
Someone suggests that we can target multiple amplicons that are nearby (say EGFR exon1, exon2 and exon3) and use them as sliding windows(fragments). Then calculate their depth on both patient cfDNA and healthy cfDNA and apply t-test.
Still, I don't think this method is valid due to lack of sensitivity (too little tumor DNA fragment in cfDNA).
Am I making sense and is there any other method to do this? Any comments are appreciated.
This has been done and by the higly regarded Schwarzenbach group: Copy number variations of circulating, cell-free DNA in urothelial carcinoma of the bladder patients treated with radical cystectomy: a prospective study. They used MLPA, from what I can see.
You appear to be relating to a NGS-based method in your post (?), but NGS is a messy and sub-standard technology. There are many other mature technologies that are more suited to the task at hand.
Yes. But I am particularly wondering if this can be done by NGS. My guess is no.
It can be done but there are large biases to deal with. Previous data that I published shows that you can faithfully detect circulating tumour DNA (ctDNA) down to a 1% dilution in blood (i.e. 1% tumour : 99% healthy cfDNA). We did that by specifically focusing on large EGFR amplifications.
The best way is to get some ctDNA and confirm the copy number of your loci / locus with a proven method, like RT-qPCR. Then, you can benchmark your NGS algorithm against this.
Can you share the publication? Really appreciate.
I've actually got my wires crossed a bit.
First, I direct you to Figure 1 of Detection of HER2 amplification in circulating free DNA in patients with breast cancer - this figure was produced by me and the data also. I show how (using RT-qPCR), one can faithfully detect HER2 ampliication in cfDNA even down to a 250x dilution. That was back when I was still a wet-lab scientist.
Second, I direct you to Influence of Plasma Processing on Recovery and Analysis of Circulating Nucleic Acids - this was actually mainly geared toward micro RNAs; however, Figure 2 tentatively shows how one can easily infer amplification (of HER2) in cfDNA.
Other data from that same group and analysed by me (unpublished) shows how one can easily detect tumour-matched SNVs in cfDNA at frequencies <5%. It was at that point that I left for the USA bu still work with the group; however, I am not up to speed with everything that they have done since. I believe that other groups have worked on this area but cannot confirm.
Determining the tissue of origin of cfDNA is also now becoming of interest via nucleosome spacing, i.e., if you detect cfDNA with cancer mutations as part of screening, you can then infer the organ/tissue from which it derived and, thus, where a tumour may be forming.
You got me scrambling to look over the vast amount of data that I had generated on that project.
I also just found this, a known PDGFRA deletion observed in tumour and here found in matched cfDNA at ~17% frequency:
This was difficult to detect, as you an imagine (in particular because cfDNA fragments are typically short), but was achieved using GATK version 2 and required the IndelRealigner.