Algorithms for calling gains and losses in array CGH data

Unbalanced 3;22 translocation with 22q11 and 3p deletion syndrome

This report describes a 25-day-old Turkish boy with unbalanced 3;22 translocation that includes the 22q11.2 deletion and 3p25 deletion syndrome. The karyotype was 45, XY,der(3)t(3;22)(p25;q11),-22. Although no immunological dysfunction could be demonstrated, the boy presented some manifestations of DiGeorge anomaly (DGA), which has been associated with monosomy for the same region of chromosome 22, velocardiofacial syndrome (VCFS), and the 3p deletion syndrome. Clinical features include short stature, hypertelorism, low set ears, cleft lip with cleft palate, short neck, truncus arteriosus, micropenis, clubfoot, over riding toes on right foot, four digits on left foot and growth delay. In addition he had feeding difficulties, respiratory infections, and developmental delay. Fluorescence in situ hybridization (FISH) studies confirmed loss of the proximal DiGeorge chromosomal region (DGCR). Array CGH analysis showed the deletion sites on chromosomes 3 and 22. This report documents a rare chromosomal aberration that causes the 22q11 and 3p deletion syndrome simultaneously.

MixHMM: inferring copy number variation and allelic imbalance using SNP arrays and tumor samples mixed with stromal cells

Background

Genotyping platforms such as single nucleotide polymorphism (SNP) arrays are powerful tools to study genomic aberrations in cancer samples. Allele specific information from SNP arrays provides valuable information for interpreting copy number variation (CNV) and allelic imbalance including loss-of-heterozygosity (LOH) beyond that obtained from the total DNA signal available from array comparative genomic hybridization (aCGH) platforms. Several algorithms based on hidden Markov models (HMMs) have been designed to detect copy number changes and copy-neutral LOH making use of the allele information on SNP arrays. However heterogeneity in clinical samples, due to stromal contamination and somatic alterations, complicates analysis and interpretation of these data.

Methods

We have developed MixHMM, a novel hidden Markov model using hidden states based on chromosomal structural aberrations. MixHMM allows CNV detection for copy numbers up to 7 and allows more complete and accurate description of other forms of allelic imbalance, such as increased copy number LOH or imbalanced amplifications. MixHMM also incorporates a novel sample mixing model that allows detection of tumor CNV events in heterogeneous tumor samples, where cancer cells are mixed with a proportion of stromal cells.

Conclusions

We validate MixHMM and demonstrate its advantages with simulated samples, clinical tumor samples and a dilution series of mixed samples. We have shown that the CNVs of cancer cells in a tumor sample contaminated with up to 80% of stromal cells can be detected accurately using Illumina BeadChip and MixHMM.

Availability

The MixHMM is available as a Python package provided with some other useful tools at http://genecube.med.yale.edu:8080/MixHMM.