bachi55/massbank2db

Build a local SQLite Database from Massbank.

massbank2db's Issues

Add function to output the MBSpectrum as MetFrag or CSI:FingerID input

See title. We need a function that can output the MBSpectrum (class) as input text-file for MetFrag or CSI:FingerID.

Add the github workflow to run the unittests automatically

...

Generate new accession title when spectra merging

When multiple accessions (spectra / MassBank records) are merged, we need to generate a new accession title and give it as property of the returned MBSpectrum object.

We can use the procedure from _to_metfrag_output:

ds = re.compile("[A-Z]+").match(self.get("accession")[0])[0]
_base_fn = ds + sha1("".join(self.get("accession")).encode('utf-8')).hexdigest()[:(8 - len(ds))]
# e.g. AU3a1fd8

Allow building a database without tandem mass spectra (MS2)

It might sometimes be useful to build the MassBank database only with retention time (RT) or simply ignoring that some records do not have an MS2.

That needs to be tested.

Move MassbankDB class parameters to the dataset insertion function

The following parameters should be moved to the dataset insertion function:

only_with_rt=True
only_ms2=True
use_pubchem_structure_info=True
exclude_deprecated=True
min_number_of_unique_compounds_per_dataset=50
pc_dbfn=None

Those are only relevant, when data is inserted. Furthermore, the MassbankDB class only acts as a wrapper around the SQLite DB file. It provides standardized functionality to access the data in the DB. It, however, does not (need to) store information about the state of the SQLite DB.

Return merge spectra

Iterating over a dataset currently returns a list of spectra per iteration, if the option to group the spectra is set to True. The list of spectra can correspond to, e.g., multiple collision energies.

We should add an option to return merged spectra, as those are required by some downstream functions like MetFrag.

Options to add merging:

• xcms implements the function mzClust_hclust using hierarchical clustering to merge the peaks in a spectrum
  -> take the function (C) and run it from python
  -> Hierarchical clustering implemented from this publication: https://link.springer.com/article/10.1007/s11306-006-0021-7
  - CSI:FingerID implements a merging function as well
  -> reimplement in python
  - implement own merging function

Test the library on the latest MassBank release

...

Test without PubChem Update

Check what happens if skip the PubChem update step.

• We cannot use the CID as row-identifier (primary key) in the molecules table
• Which information we currently assume to be added to the MassBank DB, but they require PubChem?

Implement filter for early eluting molecules

Implement a record filter based on the (estimated) column-dead-time to filter out early eluting / non-retaining molecules.

Questions here:

• Is my definition of "early" eluting molecules actually a universally applicable definition?
• How do we deal with datasets for which we cannot estimate the column-dead-time because the necessary meta-information is not provided?