Hacking Session: Draft computational chemistry ontology

[StroemPhi]

context

We could actually create the first draft of a computational ontology, based on the needs defined in the talk by Mario Wolter from last year. Notes from 2025 breakout session to this: https://docs.google.com/document/d/17xbqr4l1CeN2LNBhwCPVzo4NDkdAq4wuqEipfE7Ng-Q/edit?tab=t.0#heading=h.7cvmh7nlrx2k

expected outcome of the workshop

• first draft of such an ontology

prerequisites issues/tasks

• get a table of terms with definition
• fill out parent column of that table by asking: what entities/things need to be described?, which relationships between these things need to be described?, which characteristics/attributes of these things need to be described?)
• have at least three datasets that are to be described with that ontology
• Index COB into 4Chem TS
• evluate the existing ontologies (OntoCompChem), gap analysis

implementation issues/tasks

• use predefined table of terms to align with COB as upper level (also a training exercise to understand COB/OBO/BFO based approach better) --> Ask classification questions: is it a process, an information, a characteristic, a material entity? (We’ll only align with COB for the first draft to not waste time)
• use the TS to find existing appropriate terms to be imported (may be prepared with proposals from us to be discussed) → a side quest for those who wanna dive in deeper
• seed an ODK repo and show how to import COB and RO core
• generate SPARQL representations of the most important 5-10 competency questions → to allow CI/CD for progress evaluation

[cthoyt]

some ideas with discussion with @MarioWolter yesterday:

1. make a repo under nfdi4chem
2. make a once per month ontology hacking meeting
3. first concrete goal: curate all basis sets (then consider what kinds of "ontologization" might be needed between them. this is so they can annotate their experiments' metadata with reusable ontology terms
4. second goal: map to the World Avatar computational chemistry ontology for backwards compatibility
5. later goals: cover other parts of World Avatar ontology

see also https://leopard.tu-braunschweig.de/servlets/MCRFileNodeServlet/dbbs_derivate_00057747/ontologies_leipzig_2025.pdf

[cthoyt]

Initial work on the ontologizing basis set exchange: https://github.com/cthoyt/basis-set-ontology-demo

[StroemPhi]

We should also find out more on what Charles Bo's team has been doing in this regard in the https://www.iochem-bd.com project. He presented a part of it in the 3rd ntologies4Chem workshop: https://www.youtube.com/watch?v=cTv6C3n1NNI&list=PLlTKDYkC1Ls9wZYRUKW0b743EMOkbWucD&index=20. But the ontology was on the one hand aligned to OntoKin/OntoCape and only very small, rather under-documented and had issues regarding its publishing (no PURL etc.).

But this is definitely which we shoud considere here as previous work to reuse or collaborate if possible.

[MarioWolter]

New PhD student working on the ontology on our side: @CodoRob

[cthoyt]

should we coordiante?

[CPI-Darren]

Are my efforts of any interest to this group?

I've been trying to do two things (i) build an ontology to track all molecular modelling computation runs on a project.

I based this work on the GNVC ontology which NFDI4Chem have archived. I found deficiencies in the ontology so I only used it to extract the terms I needed. Once that was sufficiently complete I decided to (ii) try and bring the GNVC archived ontology up to date.

Caveat - subject matter expert rather than knowledge engineer.

[StroemPhi]

Your efforts are definitely of interest here and the workshop should be used to bring all relevant stakeholders together. So let's find out how we can reuse you work in this context.

[StroemPhi]

I'm wondering how we can get your input (@CPI-Darren) integrated here?
Having looked at your above linked work, which I have a real hard time grapsing based on the provided documentation, I suggest we need to take a few steps back and first rethink/state what we need the ontology to do / be able to express. I'd advocate to work data-driven, which means having at least three datasets (which could be dummy ones) that serves as examples on which we can test our competency questions. The latter presupposes to have clear competency questions written down, which is very critical not just in terms of allowing us to test the ontology iteratively, but also to communicate the subject matter to others, who, like me might not be familiar with the computational cehmistry domain at all.

Another thing I noticed is the use of the Gainsville Core Ontology in your project, but also its use in the OntoCompChem project. What is not clear to me regarding this ontology, is how mature it was, before it was discontinued. I would not encourage to use it in the present, simply for the reason it is not mainteained anymore and because its PURL is not resolving.

[CPI-Darren]

As preparation for the workshop I will work-on

1. improving the documentation;
2. provide links to data sets (eg caa) with others to come; and,
3. share the competency questions.

Also, I want to share my learnings ... eg, one (big) mistake ... the competency questions should be done at the beginning rather than at the end ... the graph (mistakenly) contains both data rather than just links to data. I am currently addressing these errors but I don't want to pretend that they didn't arise and I think it will be of value to others to share them.

The Gainsville Core Ontology was used because its authors are pioneers in the molecular modelling community. But so far I have found many gaps in it. But again, it was and continues to be another learning experience.

[CPI-Darren]

One thing I'm thinking about ... all my efforts are from the perspective of a subject matter expert not an ontologist / knowledge-engineer. I'm quite happy for everyone to use my efforts - if appropriate - as an example of how not to ... @StroemPhi has already pointed out some gaps and omissions which I've tried to address but I'm sure that there will more ...

[StroemPhi]

Your subject matter expert perspective is definitely needed! Especially with regard to the formulation of the compentency questions, as to what do you expect the ontology to be able to answer/represent. But also with regard to your thoughts about Gainsville Core and the other ontologies you might have already looked at. We wouldn't want to reuse/import Gainsville Core, since it is not maintained anymore. But its classes and relations you consider relevant, and that do not yet have an equivaltent in an active ontology of our canon, are exactly those we need to identify and then either mint in this new computational chemistry ontology, or request in an existing one like CHEMINF or RXNO.

[hfliegl]

I find this topic very interesting. However, the purpose of the anticipated ontology is not clear. From the notes of the last workshop we write "I want to describe a theoretical chemistry dataset with standardised terms." in the talk the connection to ELNs is referred to. After clarifying the purpose we could work on competency questions.

[StroemPhi]

I've created issue #21 that is intended to help the session chair better structure the hacking session. Please keep using this discussion for brainstorming before the workshop to add any comments, thoughts, links and tasks you thing should be discussed / worked on during the session. We'll try to destill this then into more actionable items within #21 shortly before the workshop.

[StroemPhi]

Thank you @CPI-Darren, for the link to the example dataset yo uprovided in #21 (comment)! Can you please use this discussion thread for further ideas, comments or links before the workshop. The corresponding issue is meant to be used during the workshop to structure the session. So my idea was to keep the discussions open for brainstorming and then the session chairs can condense this into more actionable subissues using the issue. Because here it is easier to have multiple threads, with one level of nesting/answering to posts, which keeps things a bit more clear 😉

[CPI-Darren]

Ah ... right ... I understand now ...

[CPI-Darren]

There were a number of points that I wanted to discuss that arose in #21
(unfortunately I misunderstood @StroemPhi 's note and polluted that thread with one of the points) here is one of them in the correct place ... (I hope)

"[H]ave at least three datasets" ...

I have one available here for the reaction of chloroacetaldehyde with water to form the hydrate (details, reaction scheme, and GIF of the reaction) with input, output and data folders all populated with files. It's real data therefore messy.

I am in the middle of repeating the methodology on the same reaction but with acetaldehyde and water (rather than chloroacetaldehyde) to give the corresponding hydrate. This should be a cleaner dataset when it arrives.

If anyone wants to suggest a third dataset I will try and build one, time permitting.

These could be used to test the final ontology.

[StroemPhi]

It's the first time we're using GitHub for the organization, so documenting this workflow is very agile and I didn't expect some of you being so quick 😆. So thank you for reposting here. I'll thus delete your comment in #21.

[harivyasi]

Hi everyone. I will be joining in the session this year.

Who am I

Currently I am involved in development and distribution of Chemotion ELN and one of my tasks is to bring compatibility for the .eln file format to the ELN. Building on existing efforts, .eln file, as coming from Chemotion ELN, will contain extensive ontological markup. A public release this work is scheduled before the workshop.

Why am I joining

Chemotion ELN was initially targeted towards capturing experimental chemistry workflows but has since been expanding into other domains. Theoretical Chemistry as a domain is one of our top priorities. In this workshop, I would like gather ideas and information (and generally ask questions) to understand how the new ontology can be worked into Chemotion ELN.

What I bring

Mainly (slightly outdated) field experience: In my previous career as a PhD/post-doctoral researcher, I worked on simulation of adsorbed molecules. Therefore, I have simulated isolated molecules as well as periodic structures, worked with of the tools of the trade and (of course) dealt with the complex pipeline and data management issues faced by a standard computational researcher. My effort would be to encourage that the ontology that gets shaped somehow encapsulates (i.e. invents/connects to/inherits) also common materials' science concepts (such as those related to defining a periodic system, or defects). On the other hand, I am a newbie as far as my understanding Ontology (as a domain of research) is concerned.

[CPI-Darren]

Hi All

I hope that the following helps for next week ... some of these were completed and report but as our momentum is gathering I wanted to collate and share them all in one place ...

As promised in September, prep for the session:

Start here: README (“At a glance”)
Competency questions: 16, each with a tested SPARQL query, plus a “Not answerable yet” list. Items 3–6 there are where a shared ontology would help most.
Three datasets: rem01 (small demo), aa (acetaldehyde hydration, checked against experiment) and caa (chloroacetaldehyde, a full messy project). Query them in the browser: playground
Every Gainesville Core and project term the data uses (52) as an SSSOM worksheet. 28 are original GC; 10 were added by my fork in GC’s namespace (SaddlePoint, IRC, ReactionPath, the thermochemistry properties…), and those are probably the best candidates for the new ontology.
Lessons learned: the “how not to” list.

[cthoyt]

hi @CPI-Darren, here's some commands you can use to work towards making your SSSOM file valid:

$ curl -OL https://github.com/Darren01/ont_mm/raw/refs/heads/main/mappings/ont_mm-gc-to-canon.sssom.tsv
$ uvx --with click sssom-pydantic format ont_mm-gc-to-canon.sssom.tsv

I sent a PR (Darren01/ont_mm#1) with a few initial fixes. Right now, it doesn't conform enough to the SSSOM specification for me to get an idea of how it might be used, so i'd suggest running the command and following up on its feedback

[Darren01]

Thanks Charles. Merged your PR and filled in predicate_id/object_id for every row; the file now passes sssom-pydantic format cleanly: ont_mm-gc-to-canon.sssom.tsv. 15 rows are real mappings (COB, ChEBI, IAO, OBI, CHMO, BFO, SIO). The other 38 are sssom:NoTermFound: nothing in CHEMINF, CHMO, OBI, IAO, COB, SIO or OntoCompChem covers the calculation types, basis set, method, thermochemistry or constraints. For this session, I think those 38 are the useful part, a gap list for the new ontology. One finding along the way: OntoCompChem reuses GC’s own IRIs (e.g. gc:GeometryOptimization) rather than defining its own.