Date

2022-10

Author

Acun, Melisa
Çetinkaya, Arda

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

260
views

87
downloads

Cite This

Copy number variations (CNVs) are >50bp structural chromosomal variants that represent a regional change in the normal diploid (2 copies) copy number (CN) of genomic regions. All CNs in a single cell are bound to be integers, however a population of cells with distinct subpopulations having different CNs may acquire non-integer copy number values for the total cell population. This is called genetic mosaicism and CNVs that result in such mosaicism are referred to as mosaic CNVs (mCNVs). mCNVs are encountered when a CNV is not germline but acquired later in life. Such acquired changes are frequently found in human cancer tissues and rarely in congenital genetic disorders. Determining the fraction of mosaicism (fm) is crucial in establishing disease severity, evaluating disease progression and response to treatment in individuals with cancer. Although some specific clinical genetic tests are available for determining fm of common cancer-associated structural variants, versatile methods for detecting fm for rare or novel mCNVs are yet to developed. Microarray has emerged as a commonly employed and reliable genome-wide method for detecting human CNVs which are usually undetectable by classical cytogenetic approaches. Here, we present a computational tool developed in R which we name as Copy-fm (Copy number variation – fraction of mosaicism), to address the need for detecting fm for large CNVs, using data obtained from SNP microarrays. The approach utilized by Copy-fm makes use of B Allele Frequency (BAF) data (Figure 1), one of the two fundamental values obtained from SNP microarrays for each oligonucleotide probe, the other being LRR (log2 R Ratio). Copy-fm algorithm relies on fitting cumulative distribution function (CDF) of heterozygous BAF values of given genomic regions in a sample suspected to harbor mCNVs to mCNV CDF models calculated from a set of control microarray data. The algorithm, then evaluates the goodness-of-fit by Kolmogorov-Smirnov (KS) Test to find the best fit. The algorithm of Copy-fm also tests several features in control and test data which would lead to failure of analysis or miscalculation of fm values (Figure 1). To determine the reliability of Copy-fm, we initially employed a series of experimentally generated CNV loss models for X chromosome with varying fm values using DNA samples from a mother-son pair. The fact that sons and mothers share a common X chromosome, and sons naturally have full deletion of a copy of chromosome X allows for preparation of any desired copy number value between 1 and 2 for X chromosome by mixing two samples. Using these models, we tested microarray data obtained from two commercially available platforms (Affymetrix CytoScan Optima Array and Illumina Infinium Human CytoSNP-12 v2.1 BeadChip) to compare real fm values with those determined by our algorithm. Copy-fm was able to call fm values for all of 11 Illumina-array-generated experimental models within a margin of 5% uncertainty (maximum deviation was 4.4%). However, the success of Copy-fm was lower for Affymetrix array generated data which predicted 7 of 11 experimental models within a margin of 5% uncertainty (maximum deviation was 9.3%) (Figure 1). The error margins were higher for lower fm values in both platforms. Furthermore, we tested Copy-fm using microarray data from real clinical peripheral blood samples that belongs to an individual with myelodysplastic syndrome containing a subpopulation of blood cells with two distinct and colocalizing mCNV loss regions on different chromosomes (chr5: 142,310,899–154,530,330 and chr12: 91,865,761–95,215,021). Affymetrix Optima data obtained at different time intervals revealed similar fm values for chromosome 5 (56.3%, 59.8%, 61.3%, respectively) and chromosome 12 (58.2%, 59.8%, 61.7%, respectively). As expected, the two distinct mCNVs had similar fm values at each time point (Figure 1). These results demonstrate that Copy-fm is successful in determining fm for loss mCNVs both in experimentally set mCNV models and clinical data within acceptable margins of uncertainty. In addition, fm values from example Affymetrix data sets (https://www.thermofisher.com/tr/en/home/life-science/microarray-analysis/microarray-data-analysis/microarrayanalysis-sample-data.html) for mCNV gain and loss are in agreement with those calculated by Copy-fm (Figure 1). Minimum sum of residuals has previously been used in a similar approach for calculating fm (PMID: 22277120), but usage of KS test for Copy-fm additionally provides confidence intervals for better evaluation and comparison. Without any user interference Copy-fm is able to consider loss-of-heterozygosity (LOH) and germline CNV status of a genomic region under evaluation which may lead to erroneous fm calculations. This provides an invaluable improvement for fm calculations as LOH regions in both control and test data are more common in highly inbred populations like Turkey. The approach put forward by Copy-fm to calculate fm for mCNVs is platform independent and can easily be adapted for next generation sequencing data. With adjustments, it can be utilized for genome-wide screening of mCNVs and calculating fm for uniparenteral disomy mosaicisms. Especially in cancer genetics, fm of frequently encountered mCNVs calculated by specialized locus-specific methods are being widely employed. With Copy-fm we offer a method for harnessing the mosaicism information from less frequent mCNVs encountered in microarrays, which can be utilized for uncovering unknown mCNVs, monitoring disease progression. This work was supported by TÜBİTAK (319S062) within the RiboEurope consortium and Hacettepe University Scientific Research Projects Coordination Unit (THD-2021-19532).

URI

https://hibit2022.ims.metu.edu.tr/
https://hdl.handle.net/11511/101330

Collections

Graduate School of Informatics, Conference / Seminar