If you need any more evidence that our two papers were -at least in my totally unbiased opinion- the obvious highlights of the whole Special Issue, just observe that Alexander Davis and Nick Navin chose us to write a Research Highlight about. They conclude:
SCITE and OncoNEM are innovative statistical methods that address a crucial problem in the construction of phylogenetic lineages from [single cell seq] data by using error models. These methods were shown to improve the accuracy of clonal lineages and mutational trees, as demonstrated in simulated datasets and human tumor samples.
Here is a quick overview of the type of trees OncoNEM and SCITE are looking at:
Most helpfully, Nick has also outlined our work for the next years:
Removing the infinite sites assumption will be an important next step, which has already been addressed in classical phylogenetic methods such as maximum parsimony.
Another important step will be to incorporate single-cell copy number and LOH into the tumor lineages, which can lead to missing mutations in the genotype matrix and violate the perfect phylogenies.
Another avenue of progress will be to improve the probabilistic error models. Both methods currently assume a fixed probability of error at every site, but could instead assume higher error probabilities for lower-confidence mutation calls.
Number 3 looks the easiest to me. Number 2 is what we are working on already. For Number 1 I need a bit of a think.
Interestingly, at the end of the Research Highlight Joseph Felsenstein gets acknowledged for ‘useful discussions’, so I guess he must have taken at least the briefest of looks at our stuff. And the fact that Davis and Navin still wrote a very positive review indicates to me that Felsenstein wasn’t bursting with laughter about the silly ways the cancer folks are trying to reinvent the phylogenetic wheel. That’s good news, I’d guess.