make slides for ASC Spring

StroemPhi / NMRspec

Prototype to semantify NMR spectroscopy research data

MIT License

1 stars 1 forks source link

Open StroemPhi opened 2 years ago

StroemPhi commented 2 years ago

Issue to collect the talking points.

StroemPhi commented 2 years ago

prelimineries:

to be said in talk

only jdx parsing
- plenty of trouble do to multiple dialects, --> more to this problem of getting good test data with enough fields

StroemPhi commented 2 years ago

Things to consider in the parsing process:

the Havard Dataverse data provides links to PubChem and ChEBI for the identification of the sample
- however the entities are not 100% the same
- e.g. D-Limonene dataset has links to D-Limonene = pubchemCID:440917 and to Limonene = CHEBI:15384
- assuming the PubChem entry being more correct also taking the stereochemistry into account, we're using the PubChemCID as the ID and PubChem's metadata (formula/SMILES etc) whenever possible

StroemPhi commented 2 years ago

when taking the matadata from pubchem for the sample, we're using the canonical SMILES for the smiles value

StroemPhi commented 2 years ago

Problems with JDX in the wild:

different writing of jdx keys means extra effort in parsing (= more if-statements)
- ##.SOLVENTNAME= & ##.SOLVENT NAME=
- ##.PULSESEQUENCE= & ##.PULSE SEQUENCE=
- Instrument: & ##SPECTROMETER/DATA SYSTEM
  - the former not even being a proper jdx key