If you're looking at a driver mutation, you can use the very rough back-of-the-envelope comparison to the estimated tumor purity. E.g., if a driver deletion is being reported at ~20% variant allele fraction in a sample that a pathologist determined was 40% tumor, this suggests the deletion affects one copy of a two-copy gene. However, in tumors this is pretty complex. First of all, that rough estimation only works for drivers or early passenger mutations. Secondly, since tumor genomes are highly unstable, you would need to integrate this information with data on copy number at the locus of interest - what if the gene of interest is 3n or 4n? How many alleles are affected? Thirdly, this assumes that you will get a ratio of sequence reads of the wild type and mutant copies that reflects the actual ratio, but sequencing can be biased (e.g. maybe the deletion is less likely to be sequenced for whatever reason).

In general, it is more likely that a somatic mutation in a tumor is heterozygous than homozygous, just because mutation events are rare enough that most will be changing a homozygous sites to heterozygous.