kathyxchen

Train and evaluate without `evalute_model` initialization in config does not work

4

In previous versions of Selene, we allowed users to automatically train and evaluate in 1 step without having to specify both the `TrainModel` and `EvaluateModel` classes. This functionality has somehow...

From discussion with @aaronkw & others. This will be an optional parameter that results in added column outputs for TSV/HDF5 labels file in variant effect prediction. Users can specify whether...

We would like to add a method to the API (that can be used through Selene's CLI) that can help user create a model directory formatted correctly for export to...

PR #58 allows users to use the `selene_sdk.samplers.file_samplers.MatFileSampler` for models that predict on continuous-valued outputs (e.g. regression models). Before this PR, the MatFileSampler only supported integer labels, and all other...

Currently, we only assign labels to the center "bin" of the sequence. With larger sequence contexts, we may want to retrieve the labels for multiple bins in a sequence to...

This is needed to begin supporting more RNA and protein sequence models.

speed up in silico mutagenesis

1

Add code for in silico mutagenesis and test cases. This method should be able to accept an input sequence and output sequences where N=1 number of bases have been mutated...

Adding an RNA sequence class?

1

This is something I'd like to consider implementing, in the hopes that it would be used in one of the examples for the paper. It might not be necessary, but...

A user should be able to get a list of important "subsequences" for an input sequence for every genomic feature. These subsequences will be built based on the mutated base...

If there is strand information for the genomic features dataset, we should be able to handle this. We may add a parameter about strandedness (this relates to the "wrapper non...