Global-Chem: A Free Dictionary from Common Chemical Names to Molecules | Hacker News Reader

Global-Chem: A Free Dictionary from Common Chemical Names to Molecules

globalchemistry.org

35 pointsSulstice10 comments

We are writing the molecules to common chemical names that the general public use and keeping a record for all of us to download and maintain together. We can all manage it as an open source governing community and use tools to declare whether these chemicals are safe or not for general use.

I'm gathering information currently, and looking to expand to other open source software tools. Does anyone want to join, have an idea for a connection, or a chemical list they think would be worthwhile.

It was just mentioned to me I should look into pre-workout supplements for example.

There's a mapping of various chemical identifiers/name to PubChem compound id at ftp://ftp.ncbi.nlm.nih.gov/pubchem/Compound/Extras/ . Quoting Quoting README-Extras in the that directory there are three files you might look at:

  SID-Synonym.gz:

    This is a listing of all (live) SIDs with their depositor
    provided synonyms (one per line). It is a gzipped text file with
    SID, tab, synonym on each line.

  CID-Synonym-filtered.gz:
  CID-Synonym-unfiltered.gz:

    These are listings of all names associated with a CID. The
    unfiltered list are names aggregated from all SIDs whose
    standardized form is that CID, sorted by weight with the "best"
    names first. The filtered list has some names removed that are
    considered inconsistent with the structure. Both are gzipped text
    files with CID, tab, and name on each line. Note that the
    names may be composed of more than one word, separated by spaces.

A similar list may be available from ChemSpider or eMolecules - I don't know.

https://wikipedia.cheminfo.org/smiles.txt is a mapping from Wikipedia title to SMILES, done using the chemical infobox.

> We can all manage it as an open source governing community and use tools to declare whether these chemicals are safe or not for general use.

Yeah, that's not going to happen.

You're in for a world of hurt if you try to crowdsource this list and find lots of contributors.

As the old phrase goes, "The dose makes the poison". Oxygen can be toxic - https://en.wikipedia.org/wiki/Oxygen_toxicity .

The California Prop 65 list ("Safe Drinking Water and Toxic Enforcement Act of 1986") includes "Ethyl alcohol in alcoholic beverages" as a chemical "known to cause cancer, birth defects or other reproductive harm" - https://oehha.ca.gov/proposition-65/chemicals/ethyl-alcohol-...

Opinions vary, sometimes with strong emotions. In this project, who has the final say on if thiomersal is safe or not for general use? (For the matter, what does "general use" mean?)

I guess that's something I am really trying to figure out. The project is moving into the Linux Foundation on Non-Profit Community Platform. Contributors have so far been from a variety of universities across the world and have been contributing in.

I am still trying to figure out the mechanisms as to how to create this type of government mechanisms. I have a couple of advantages over the other platforms I think. Mine is organized in such a way data can be found easily and categorized. My lists are abstracted from expertise who and what resource so folk can trace easily. I also wrote a lot of tooling for analysis and other things you can use to analyze the data yourself to get an idea of the functional groups and diversity. Things that relate to your compound.

I didn't see the Prop 65 list and that's very cool.I think that is something I need to keep watch of as well.

> Mine is organized in such a way data can be found easily and categorized.

I do not know what this means.

It's very easy to find compounds in PubChem or ChemSpider.

I didn't see how to find, say, 1,3,7-Trimethylxanthine in your system.

As far as I can tell, it requires using the no-cost version of ChemDraw to convert the name into a structure and then ... do what?

> My lists are abstracted from expertise who and what resource so folk can trace easily.

How is this different than what ChemSpider does? They also offer "Manual comment and correction tools offer crowdsourced curation with expert review" and "You are invited to help build this community for chemists by contributing your structures, spectra, or syntheses. You can also actively participate in the curation of the data; registration enables you to become a depositor or curator." (quoting https://www.chemspider.com/AboutUs.aspx ).

Is the main difference that you organize things by lists?

In that case, couldn't you save a lot of effort by being an overlay into ChemSpider? Let them handle the chemical entry, name validation, property prediction, and so on.

> I also wrote a lot of tooling for analysis and other things you can use to analyze the data yourself to get an idea of the functional groups and diversity.

I didn't see any of that tooling from your site.

What is your measure of "diversity"?

Could you elaborate on your internal naming scheme for compounds? For example, https://github.com/Sulstice/global-chem/blob/ec4a5cb28a6360c... has:

            'haloalkane': 'CBr',

That should be bromomethane, yes? "Halo" is any halogen, and "alkane" is any alkane, so "haloalkane" should refer to a class of compound, yes?

And a few lines down,

            'nitrile': 'N#CC',

Here again, "nitrile" is a class of compounds - the specific compound is acetonitrile.

Picking another data set, https://github.com/Sulstice/global-chem/blob/ec4a5cb28a6360c... has:

             'detla-9 tetrahvdrocannabinolic acid c':
             'CCCCC1=CC2=C(C3C=C(CCC3C(O2)(C)C)C)C(=C1C(=O)O)O',

That should almost certainly be "delta" and "hydro", yes?

And line 427 of that file has:

           'zanthophylls': 'CC1=C(C(CC(C1)O)(C)C) ...

when I think that's 1) supposed to be "xanthophyll", which 2) is a class of compounds, and 3) PubChem tells me the compound is "beta,epsilon-Carotene-3,3'-diol, (3R,3'R,6'R)-", see https://pubchem.ncbi.nlm.nih.gov/compound/181579 .

Also, it appears "constituents_of_cannabis_sativa" is a subset of what's available from https://pubchem.ncbi.nlm.nih.gov/taxonomy/3483#section=Natur...

It would be helpful if you provided a how to guide for new contributors.

Eg: say I was willing to go through all the cleaning products or supplements in my house and submit lists of what's in them, how would I go about doing that?

That's a good idea and I should add it to my list of stuff to add too. I have wondered how to get that type of feedback mechanism working.

Do you have a plan to release a web-based UI like Wikidata so the data could be more accessible to the general public?

Currently, i'm doing this exact same things (identifying chemical names found on products) on wikidata in my pastime.

Step 1 makes sense, but:

> Step 2: Create a Governing AI that we work with

> True General Artifical Intelligence comes from a variety of people and not one. Our AI is based off the data the people report to help us understand what we consume and use so we can understand medical conditions more and learn how to diagnose easily.

What is this all about? Sounds vague to the point of meaninglessness.

TIL about #cheminformatics and Linked Data (Semantic Web):

Cheminformatics: https://en.wikipedia.org/wiki/Cheminformatics

https://github.com/topics/cheminformatics :

- https://github.com/hsiaoyi0504/awesome-cheminformatics #Databases #See_also

- https://github.com/mcs07/PubChemPy :

> PubChemPy provides a way to interact with PubChem in Python. It allows chemical searches by name, substructure and similarity, chemical standardization, conversion between chemical file formats, depiction and retrieval of chemical properties.

http://chemicalsemantics.com/introduction-to-the-chemical-se... ... /? chemicalsemantics github ... https://github.com/semanticchemistry/semanticchemistry #See_also :

> The Chemical Information Ontology (CHEMINF) aims to establish a standard in representing chemical information. In particular, it aims to produce an ontology to represent chemical structure and to richly describe chemical properties, whether intrinsic or computed.

Looks like they developed the CHEMINF OWL ontology in Protege 4 (which is Open Source). /ontology/cheminf-core.owl: https://github.com/semanticchemistry/semanticchemistry/blob/...

- Does it -- the {sql/xml/json/graphql, RDFS Vocabulary, OWL Ontology} schema - have more (C)Classes and (P)Properties than other schema for modeling this domain?

- What namespaced strings and URIs does it specify for linking entities internally and externally?

LOV Linked Open Vocabularies maintains a database of many RDFS vocabularies and OWL ontologies (which are represented in RDF) https://lov.linkeddata.es/dataset/lov/terms?q=chemical

- "The Linking Open Data Cloud" (2007-) https://lod-cloud.net/

/? "cheminf" https://scholar.google.com/scholar?hl=en&as_sdt=0,43&qsp=1&q...

/? "cheminf" ontology https://scholar.google.com/scholar?hl=en&as_sdt=0,43&qsp=1&q...

"The ChEMBL database as linked open data" (2013) https://scholar.google.com/scholar?cites=1029919691588310633... ... citations:

"PubChem substance and compound databases" (2017) https://scholar.google.com/scholar?cites=7847099277060264658...

"5 Star Linked Data⬅" https://wrdrd.github.io/docs/consulting/knowledge-engineerin...

Thing > BioChemEntity https://schema.org/BioChemEntity

Thing > BioChemEntity > ChemicalSubstance https://schema.org/ChemicalSubstance

Thing > BioChemEntity > MolecularEntity https://schema.org/MolecularEntity

Thing > BioChemEntity > Protein https://schema.org/Protein

Thing > BioChemEntity > Gene https://schema.org/Gene

Some of the BioSchemas work [1] is proposed and pending inclusion in the Schema.org RDFS vocabulary [2].

[1] https://github.com/bioschemas

[2] https://github.com/schemaorg/schemaorg/issues/1028

Will newer Bioschema terms like BioSample, LabProtocol, SequenceAnnotation, and Phenotype be proposed for inclusion into the Schema.org vocabulary?: https://bioschemas.org/profiles/index#nav-draft

Reply on news.ycombinator.com