hogfeldt / ctdnatool Goto Github PK

Make a generator that can create a tensor with the following dimensions (genome region, frag. length, end sequences).
The sequences should be found in the in the reference genome.
In order to store the sequences as counts, create a mapping from seq string to a base 4 string where the nucleotide having the following values
A -> 0
T -> 1
G -> 2
C -> 3
The index in the tensor is then the conversion from base 4 to base 10.

We should have a manipulation that can bin the first axis of a matrix or tensor, with a sliding window.
The user should be able to set the bin size and the stride.

If the bin size and stride is chosen in a way such that the last bin exceeds the array, its probably best to just create a small bin an throw a warning.

• Find a more suitable name
• Should be able to take data instead of sample file
• Find duplicates in scripts
• Add to CLI

When testing the combine_data command, it was found that some length_end_seqs samples (.pickle) cannot be read, because they are truncated.

Use sys.path for getting name of files, instead of splitting the string.

Right now data is stored in the .npy or .pickle format with a corresponding index file. It could be nice to integrate the index file in a little wrapper class, that could contain both data and meta data.

The naive way for storing such an object is in the pickle format. My only concern with the pickle format is that it is not secure see documentation (https://docs.python.org/3.9/library/pickle.html).

Maybe we should consider finding an alternative? Maybe hdf5?

Cap end of bins if they extend beyond end of vector.

Last column in data passed to summaries_data() manipulator contains sequences containing "N". This needs to be handled.

Fragments that cannot confidently be mapped to the reference genome should be filtered out.

We should have a test suite so we are sure that we don't break stuff when adding new features.

Given data and an index, the manipulator should sum the axis given by the index.

Refactor cli_flow.py to use file_of_files.

We should have a class that can handle IO and store data and meta data.

Document the commands in cli.py.

When the chromosomes are pulled from the 2bit file, then sort them so the order is consistent.

Both for debugging and for issuing warnings, it could be nice the use the logging module for logging instead of just using print statements.