When you have matched tumor-normal SNParray samples and are concerned about underestimating the copy number profiles of your tumor sample because of tumor contamination from immune cells, blood or nearby tissue you can estimate the relative proportion of tumor and normal cells in a sample (tumor cellularity) by a new tool called qpure
The authors of qpure reasoned that a SNPs with Loss of Heterozygosity, the B allele frequency (BAF) of that SNP should change from 0.5 to either 0 or 1 (Genotype change from AB to A or B). When the BAF does not completely shifts from 0.5 to either 0 or 1 the difference is an indicator of contamination with non-LOH/normal cells. We have tested the tool on ~70 tumor-normal SNParray samples and we see that it correlates well with pathology scores (the authors even claim that qpure is even a better descriptor of tumor cellularity than pathology). I think it can be very useful to get an idea whether your tumor copy number profiles are underestimated in specific samples and I recommend this tool to other people who have SNParray tumor-normal data.