search

griffithlab / pVACtools

http://www.pvactools.org
BSD 3-Clause Clear License
137 stars 59 forks source link

Have pVACvector output a list of problematic peptides #589

Open susannasiebert opened 3 years ago

susannasiebert commented 3 years ago

Ideally we would come up with a heuristic to identify the overall most problematic peptides for a whole pVACvector run if no valid paths is found. These could then be suggested to the users as peptides to remove from their candidate list.

susannasiebert commented 5 months ago

Alternatively, it would be straightforward to remove nodes from the graph after all spacers have been tested to try and get a partial solution. Running the annealing procedure is relatively compute cost effective so we could test any graph with an combination of any number of nodes removed. We may want to considering making this a user-specified upper limit because it might be acceptable to have, e.g. 3/20 peptides removed from the pool.

The algorithm would start by removing one node and test all possible graph options with a single node removed. It will output all solutions that might be found for these partial graphs. This is so that if there are multiple solutions with one node removed, the user will be presented with all the options since the removal of one candidate might be preferred over a different candidate (e.g. a tumor driver mutation). If one or more partial solutions with one node removed are found, the algorithm stops. If not it will test the graph with any permutation of two nodes removed. This repeats until the max number of allowable removed nodes is reached or a partial set of solutions is found.