Skip to content

Commit

Permalink
Merge pull request #1 from an1lam/fix-typo
Browse files Browse the repository at this point in the history
Fix typo
  • Loading branch information
an1lam authored Dec 16, 2019
2 parents 6475a32 + ef99e2b commit 24c2d26
Showing 1 changed file with 1 addition and 1 deletion.
2 changes: 1 addition & 1 deletion docs/templates/mutation_map.md
Original file line number Diff line number Diff line change
Expand Up @@ -91,7 +91,7 @@ The meaning of the parameters is identical to the ones in the python API mention

This section gives the necessary information to understand how the tested region is derived from a query file.

In order to perform a query on a model the query input must be transformed into genomic regions compatible with the model. Similar to variant effect prediction using the `score_variants` the automatically chosen region generation method will be chosen based on whether a dataloader offers a bed file input for postprocessing. `dataloader.yaml > postprocessing > variant_effects > bed_input`. By setting this value the mutation map method will automatically generate a temporary bed input file requesting model input for genomic regions. The path of this temporary bed file is then passed on to the dataloader by resetting the respective argument in the `datalaoder_arguments`. For some models it is not possible to freely define the genomic region for which model input data should be generated - in that case the dataloader.yaml does not have the `dataloader.yaml > postprocessing > variant_effects > bed_input` set. In those cases the datalaoder is executed without modifying the `dataloader_arguments`. The metadata generated alongside the model input is then used to identify model input that overlaps a query region / query variant.
In order to perform a query on a model the query input must be transformed into genomic regions compatible with the model. Similar to variant effect prediction using the `score_variants` the automatically chosen region generation method will be chosen based on whether a dataloader offers a bed file input for postprocessing. `dataloader.yaml > postprocessing > variant_effects > bed_input`. By setting this value the mutation map method will automatically generate a temporary bed input file requesting model input for genomic regions. The path of this temporary bed file is then passed on to the dataloader by resetting the respective argument in the `dataloader_arguments`. For some models it is not possible to freely define the genomic region for which model input data should be generated - in that case the dataloader.yaml does not have the `dataloader.yaml > postprocessing > variant_effects > bed_input` set. In those cases the datalaoder is executed without modifying the `dataloader_arguments`. The metadata generated alongside the model input is then used to identify model input that overlaps a query region / query variant.

For cases when genomic regions can be defined freely for a model, the input samples will always have to generated matching the model input sequence length. This means that for query variants a region of the length of the model input will be centered on the query variant position. For query regions (e.g.: bed input file) every region is overlapped with windows of length of the model input. The first of those regions will start at the same position as the selected query region. Regions of the length of the model input sequence length will then be generated consecutively in order to cover the full region defined by the respective query region - see this schematic:

Expand Down

0 comments on commit 24c2d26

Please sign in to comment.