Probabilistic approaches for profiling copy number aberrations and loss of heterozygosity landscapes in cancer genomes