1. How do I know if ichorCNA completed successfully?
2. How do I assess the data quality for my sample?
3. Does my cell-free DNA sample contain tumor-derived DNA?
4. Does my cell-free DNA sample contain sufficient tumor-derived DNA to proceed with whole exome sequencing?
5. Can ichorCNA be used on non-human ULP-WGS data?

You should be able to find the following output files correctly written: <sampleID>.cna.seg, <sampleID>.seg.txt. <sampleID>.seg, <sampleID>.params.txt, <sampleID>.correctedDepth, and a directory containing plots <sampleID>/<sampleID>_CNA_chr*.pdf, <sampleID>/<sampleID>_genomeWideCNA.pdf. Please refer to the Output section for more details on these output files.
Back to top

We recommend looking at the GC-Map correction MAD value, found in the <sampleID>.params.txt file, to assess the data quality of your sample. A value greater than 0.3 usually indicates a high variance in the data and it is likely too noisy. ichorCNA corrects for GC and mappability biases in the data but a high MAD score may indicate another source of noise or artefact that is not being corrected. To restrict your results to only “high quality” data we suggest only using samples with a GC-Map correction MAD less than 0.15 and also ensuring the mean coverage (without PCR duplicates) is at least 0.1x (not calculated by ichorCNA).
Back to top

At 0.1x coverage, ichorCNA benchmarking (Adalsteinsson et al. 2017) revealed a lower limit of detection of 0.03 (3%) tumor fraction for detecting the presence of tumor in the sample. This benchmarking was performed by in silico mixing of metastatic breast and prostate cancer patient cfDNA with healthy donor cfDNA; therefore, various types of copy number events (e.g. chromosome/arm-level aneuploidy, focal events, high-level amplifications) were considered. We also determined that a sample containing only a single arm (~100Mb) gain and a single arm deletion is required to detect presence of tumor at the 3% limit.
In practice, because your data may differ, including being a different cancer type, having different sequencing coverage, or different data quality (see How do I assess the data quality for my sample?), it is recommended to inspect some of the samples with estimated tumor fraction around 3%. Please the Output - Interpreting the results section for details.
Back to top

For exome sequencing we recommend using a Tumor Fraction (params file) cutoff of greater than 0.1 without manual review. It may also be possible to do exome sequencing for samples with an estimated Tumor Fraction between 0.05 and 0.1 but we recommend reviewing the genome wide plots for these cases and confirming that the correct solution was picked. The model tends to err on the side of underestimating tumor fraction so it's possible that some of these cases can be rescued. Please the Output - Interpreting the results section for details.
Back to top

We have not tried analyzing data from other species and this is currently not supported. If you would like to try we would suggest generating reference GC and mappability WIG files using HMMcopy for your specific genome. You will also have to modify the --chrTrain and --chrs parameters for your species of interest.
Back to top