• Open Science
• Recognition & Rewards

The provide a good number of additional pointers for important context.

Recognition & Rewards generally is a very complex question. Just thinking about how people reply to the Cornell University rape allegations. Not every researchers is evaluated in the same way.

And then there is the highly complex approach how to compare researchers when the question of promotion comes up: qualification or quantification. They both measure certain aspects of our performance, but not all metrics (qualitative or quantitative) actually measure the performance of the researcher. The journal impact factor is a clear example of that. The H-index is more personal but also does not really measure what you want to use it for. These things too return with Open science monitoring. We have to monitor such efforts critically (like we do with everything, the researchers we are).

The dynamics are fascinating. But they are also essential to my work. I am line manager of multiple researchers, each with their own career paths. Under to modern HR, any Maastricht University researcher can opt to use career compass option. This career policy gives our employees more freedom how to be evaluated. For example, it gives more room for recognition for teaching and for patient care. But all routes have four obligatory topics:

• Academic citizenship
• Personal and professional leadership
• Team performance
• Impact & Open Science

In the evaluation meetings with your development board you will discuss how you did on the activities related to your personal career compass. And then again the qualifications and quantification come up. And that opens the can of worms.

Now, because open science and team science are used a lot in the Theme Bioinformatics of our Department of Translational Genomics, I like to know what options I have to recognize and reward that.

That is why I signed up for the Recognising and Rewarding Open Science programme that started last week.

Before we go there, just a quick note on what ToxTemps are and ToxTempAssistent actually is:

The ToxTemp template, based on OECD Guidance Document 211, standardises reporting for cell-based NAMs. However, completing its 77 questions constitutes a substantial bottleneck. The aim of this study is to introduce ToxTempAssistant, a Large Language Model (LLM)-assisted web tool that supports toxicologists in drafting ToxTemp documents based on user-supplied context documents. This study quantifies the tool’s baseline performance under controlled conditions. ToxTempAssistant uses grounded, per-question prompting with mandatory source attribution.

The LLM aspects

AI is very old. Arguably, my PhD thesis had this as key topic, though I preferred to use to term chemometric or machine learning. Critical thinking has been essential to this field for a long time, and much of the PhD thesis is actually about critically assessing the performance of the methods used in the thesis. There is decades of research how you do this. Sadly, when it comes to Large Language Models (LLMs), these are not routinely used.

LLMs are indeed something new. I have seen natural language processing research when I was Cambridge with Peter Murray-Rust. The current LLMs are different, less deterministic, more probabilistic. From a chemometrics perspective, that makes sense. Language is complex, and not so deterministic in itself. Moreover, when representing words, sentences as numbers, you can integrate any resource (think data tables, images, etc). Even more, digital representation was even more the central theme of my PhD thesis.

But because of the nature of the method, the nature of how the commercial, better known models are trained (and the impact on the notion of copyright), the impact on the climate emergency, the black box that these LLMs often are, there are so many technical, scientific, and ethical reasons to stay away from them.

You can write books about that. Literally (I did not read them yet):

• The Nerd Reich
• The AI Con

(I have the feeling I am missing one title I wanted to highlight. If I remember, I will add it.)

And more here and here. And I am looking forward to reading Deep Unlearning: The Rise of AI and the Radicalization of a Tech Idealist. Also, I recommend at least following Timnit Gebru and Emily Bender.

A year ago, I signed the Open Letter: Stop the Uncritical Adoption of AI Technologies in Academia, now signed by more than 2,000 people (it is not too late).

So, with all these things mind, I am happy that Jente did critically adopt LLMs in her work. The paper includes various experiments to explore the impact of various model parameters on generating nonsense. She also explored to use of alternative LLMs platforms, opening the option that some day it runs on other, more ethical platforms. ToxTempAssistant, moreover, limits the material it takes information from, from a limited set of sources, provided by the users. Of course, the model itself is still training on resources with questionable provenance.

Jente’s paper describes the use of positive and negative controls to put the performance in perspective. In doing so, the paper formalizes how to evaluate the use of LLMs in situations where the LLM is used to summarize other reports, a common thing to do. This allows us to monitor the impact of, for example, new LLM releases.

The results are promising and various independent projects have shown interest in adoption. The ToxTemp reports are important for safety assessment: they provide essential context to experimental results and as such essential to the FAIR-ness of toxicology data.

Open Peer Review

The ToxTempAssistant paper is published in the relatively new journal Evidence-Based Toxicology (EBT):

Evidence-Based Toxicology is a broad-focus, gold open-access journal, created to support the use of open science practices and evidence-based methods in toxicology and environmental health.

So, CC-BY license (gold open access) and support for Open Science. And Open Peer Review, as we will see. Also, I understand it is not a diamond open access journal, so expect APCs.

The journal has a community on Zenodo for preprints and peer-reviews. I think this is a really nice choice. Of course, a journal specific preprint server has downsides too. For example, if it gets rejected from EBT, do you use the EBT preprint server when submitting to another journal? Do you upload a new version (after all, you should address some of the comments why it was rejected) to another preprint server?

The EBT preprint gets a record and revisions are uploaded as new versions to the same Zenodo record (doi10.5281/zenodo.17192970):

But because EBT uses open peer review, there is a parallel Zenodo entry with the reviews (doi10.5281/zenodo.17278785):

The last version here is the acceptance notice:

One of the reviewer suggestions was to use the TRIPOD-LLM template (doi:10.1038/s41591-024-03425-5), which was included in a later revision. That made a lot of sense and is a nice example how the journal actively works on applying open science ideas. The journal webpage writes:

We define “open science” as the set of practices aimed at improving the transparency, validity, reproducibility, and accessibility of scientific research, while promoting equality of opportunity to participate in and benefit from the products of said research.

One comment here is that the template asks on what page that point of the checklist is discussed in the article. That is nice, but links it directly to the revision of the manuscript that form applies too, but that is not reported in the template. Not ideal. Then again, the point is the checking, so maybe not a big deal.

Another comment is that it seems the journal website’s page for the article does not seem to actually link to the preprints nor the peer-review reports (or acceptance note). So, in time, these open science aspects of this article will likely get lost in time. Who will find those reports if the article does not cite them? This will require Taylor&Francis to modernize the publishing model, and I sincerely doubt that that will ever happen.

Prof. Anne Kienhuis, one of the co-authors, suggested this journal lead by Prof. Paul Whaley. And I am happy to have seen this new approach in action and like to thank Paul for pushing for these innovations. I recommend trying it yourself for your next toxicology work.

References

It me also made me realize something else. We have by now a good practise of tracking retracted articles, and platforms to discuss articles openly, like PubPeer. The second can be used for datasets too, when they are deposited in DOI-providing resources, like Zenodo or Figshare.

Fraudulent data can certainly result in papers that describe the data to be retracted. But retracting the publication does not mean the paper does not get cited anymore, see doi:10.1007/s11192-020-03631-1 or Scholia for retraction data. But data that starts having a live on its own after the deposit in some archive may worsen this problem. After all, our practises for sharing retraction flags and concerns are even without this substandard.

Should we not also start taking more care about the data we publish? Track concerns about them? Maybe even a way to retract datasets, when fraud was in play?

References

Many universities have green Open Access repositories, like Maastricht University and Radboud University. These repositories are scattered and contrary to FAIR principles, metadata is not shared widely. Of course, prominently, scholarly publishers could provide that information. Any paywalled journal article could have a link to a legal green open access version. It could. It would make total sense if spreading scientific knowledge is your goal. But we live in an alternate universe.

We are relying on third-party indexes. I here briefly describe two options, but love to read blog posts about other alternatives! I am not looking for routes like PubMed Central or Europe PMC that come with huge collections of open-licensed content, including in the FAIR JATS format

It’s really the non-OA-licensed green Open Access that we need to get indexed better.

Oh, and, of course, all Dutch literature must become green Open Access.

OpenAlex

Unpaywall (see also this post) has been collecting open access versions but is now including in OpenAlex. The annotation is quite rich and here the question is more, how can the open science community enrich this OpenAlex content.

Wikidata and WikiCite

Wikidata and WikiCite are projects aligned with Wikipedia, and [citation needed] is an essential part of that. We have used it in our research in multiple occassions. The curation process is transparent, providing the essential provenance needed for scientific research.

It provides the work available at URL property and it has been used to link to green Open Access versions. The extend, however, is less clear (to me). But another problem is that Wikidata does not have the ambition to be a generic solution. That said, all extra links help.

Schema.org and web search indexes?

That leaves me, for now, with the role of generic web searching. Doesn’t that help? I can just search for the article title, right? Well, yes. But the process of how to get your green open access article indexed is not clear. There are many metadata standards that your repository needs to support (e.g. OpenAlex lists 8 and that does not include schema.org) and that does not guarantee you get indexed. Worse, commercial companies are know to remove content in a opague process, like with this video about the Meta gluurbril. We cannot trust that green open access metadata gets preserved.

Maybe open web indexes will make a difference, but then we still need a simple API to integrate into our tools that just returns the information. I had a go at this last weekend. It would look something like this:

But things go slowly, and FAIR metadata most certainly. But if we consistently take steps, things do change. For example, after 20+ years, CrossRef now supports blog as research output type.

I like to stress, metadata is being shared by repositories, but we should support this a bit more. We should reuse that data, support that, curate that, etc. Let’s support our fellow open science community members! And that starts by sharing our knowledge on how all this works, does not work, could work, and what small steps we can take today and tomorrow. Looking foward to your ideas!

WikiPathways was already around five years old when I created my account. It was created in January 2007 (doi:10.1038/npre.2010.5361.1). It was first formally published in 2008 (doi:10.1371/journal.pbio.0060184), so the project is about 20 years old (maybe a new WikiPathways conference next year?). The launch made it to Science (doi:10.1126/science.321.5889.623c):

The makers of GenMAPP [..] have launched a site for sharing findings on metabolic pathways. [..] WikiPathways (wikipathways.org) offers a way to integrate information on these complex networks [..] The site has more than 300 registered users and contains information on 500 metabolic pathways in seven species, including humans.

And Nature (doi:10.1038/455022A):

Alexander Pico remembers just when the idea hit him. In January 2007, he and his boss, Bruce Conklin, were discussing how to push their software tool for visualizing intracellular signalling pathways to the next level of interactivity — when Pico blurted out, “What we really need is a wiki!”

Just like review article curate and put primary literature in context, so that WikiPathways. They pitched it at the time as modelled after Wikipedia, but surely it is also just as well modelled after traditional knowledge dissemination approaches. One can easily argue that the wiki approach is essential: anyone can create an account. And this has been held against WikiPathways. But if you look at the who actually edits WikiPathways, these are only people working in an academic research environment.

WikiPathways is, of course, used as a reference knowledgebase, and the use in pathway enrichment and networking approaches are perhaps the best known. Another aspect of a wiki, however, is the community approach: a community can work out a collection of machine-readable (FAIR) pathway models, even when the discovery is in full swing. WikiPathways is also a platform to exchange knowledge, discuss literature, do research. The use to of this feature turned out really useful at the start of the COVID-19 pandemic.

Genomes, species, and pathways

The set of full genome assemblies was not that large; in 2021 there were 3278 genome assemblies for animal species (doi:10.1073/pnas.2109019118 and the Genomes on a Tree (doi:10.12688/wellcomeopenres.18658.1) reports 27883 assemblies for 664 complete genomes.

For WikiPathways, seven species in 2008 grew to 39 species today today, though many of them only have a few curated pathways. Indeed, human is still the predominent species, in number of pathways (visualized with this SPARQL query).

That said, several communities have worked in the past to create sets of pathways around specific species. In that post, I already mentioned Wormbase, but we also have a collection of pathways from YeastPathways (doi:10.1093/genetics/iyae185), for example wikipathways:WP137. Now, the yeast pathways are not in sync with the upstream SDG YeastPathways, I think, so there is some curation to do on our side (from the 2025 Genetics paper):

As the first major update since 2012, we updated 62 pathways with expert summaries on pathway genetics, biochemistry, regulation, and more. Thirty-three new pathways with specificity for yeast biochemistry were propagated from MetaCyc at SRI (Caspi et al. 2018), and 105 existing pathways were edited for proper enzymatic classification, reaction connectivity, and gene attribution. Compounds that were previously missing a chemical structure have also now been updated, along with the stoichiometry and scheme of many pathway reactions.

In similar style, WikiPathways has had similar project around setting up a pathway collection for a new species, like cow in a project by Zahra Roudbari (doi:10.3389/fgene.2019.01370). And more recently atlantic cod (ongoing research by Madhushri S. Varunjikar from the Norwegian Institute of Marine Research).

Homology converted pathways

But WikiPathays also has a feature to convert pathways from one species to another. With more and more whole genome assemblies becoming available, so are mappings of genes between one species and another. This is the basis of that cow project (from the article):

The online biological pathway repository, WikiPathways (Slenter et al., 2017), contains pathways of different species, however a B. taurus collection was missing.

And:

Second, the WikiPathways homology based the homology mapper which is available at GitHub (https://github.com/PathVisio/homology.mapper) was updated to improve homology coverage for gene products that were annotated with different data sources. The pathways were converted from human pathways, with a required minimum successful conversion of at least 50% of the original human genes.

The question is, of course, what is the best model species to use. That decision can be based on multiple aspect, including the use case. There is freedom.

So, when this week the book chapter Xenbase: A Guide to the Xenopus Genetics and Genomics Knowledgebase was published (doi:10.1007/978-1-0716-5360-9_3), and I was pinged WikiPathways was cited, I started reading.

WikiPathways are linked via gene symbol (Fig. 6A, black arrow) link redirects to a pre-set gene symbol search on this community-curated open science resource [20].

The linking via gene symbol made me write up this post. If they have the mappings, then with the GPML Homology Mapper we can create pathways for the two frog species (X. laevis and X. tropicalis).

Species specific curation

And there are reasons to do that. WikiPathways allows you to cite literature, at a pathway level, for a gene, protein, or metabolite, and even at the level of an interaction. That literature is species specific. The curation can also work out what to do with those parts of pathways for which no genes were mapped. Having this is machine-readable knowledge has shown to be of great usability, at the very least making the knowledge much, much easier to find.

And that brings us back to another role of review articles, of book chapters giving expert overviews, but also of wiki’s: put knowledge in perspective of the literature from which it is derived. I wish pathways database, generally, are used to communicate and get peer reviewed the fruits of the work studying the biological processes on all these species.

With more than 600 full genomes and 39 species in WikiPathays, we have an exciting and growing seed.

References

I already knew that Crossref tracks citations (see e.g. this API call for a paper where Jmol is used, doi:10.59787/2413-5488-2025-49-1-83-92), and I learned that apparently Crossref also does an good job at finding citations when blog post URLs are used instead of the DOIs of the blog posts.

But how does it know? When Rogue Scholar registers a DOI, I assume it passes on the URL to redirect too. But what if the URL changed over time? For example, I used blogger.com for many years (with satisfaction) and the URLs of the posts have changed!

Crossref, how do I get the old blog post URLs also linked to the DOI?

For example, the doi 10.59350/65nqr-3w351 links now to https://chem-bla-ics.linkedchemistry.info/2023/08/18/last-post-here-freebie-model-online.html but originally the URL was https://chem-bla-ics.blogspot.com/2023/08/last-post-here-freebie-model-online.html .

References

Platforms are getting more better at precisely capturing how we communicate science. Rogue Scholar is a great example of that, like the no longer existing CiteULike before. And Rogue Scholar has been challenging my Markdown-based blog post to support its functionality. For example, I had to add a JSON Feed, but also was happy that Rogue Scholar added CiTO Support.

But capturing all citations between things remains an issue. Platforms like Google Scholar and Web of Science do their best, but their best is not Open or nor FAIR or neither. Generally, it has taken way too long for DataCite and Software Citations to be picked up. However, we can take control ourselves when solutions do not exist.

And because I want journal-to-blog citations are FAIR too, I hacked up another extension of my blog today. I can now add lines like this to the metadata of blog post:

cited-in: cito:discusses:10.1186/1471-2105-8-487

What is encodes is that the article with doi:10.1186/1471-2105-8-487 (one of my own) discusses one of the posts in my blog.

Online, this is rendered with citation.js like this:

Here, the References are the things I cite in my blog post, and the Cited in are things citing this blog post.

How blog posts are cited

Now, DOIs for blog posts are a new thing: they did not exist when I started blogging. And for the 12 citations I found so far, they are not using a DOI. I doubt that these citations are easily found in CrossRef or DataCite. Here are two examples, of increasing complexity. The first one is how many citation styles would do it:

But they can also look like this:

And scientific journals sometimes even use URL shorteners and then it may look like this:

This makes it a challenge for any platform to pick up these citations!

How about the JSON Feed

That leaves me with how I make this information avaiable, a bit towards FAIR, but at least machine readable. This is done with a JSON Feed extension (at the time of writing, I did not add this proposal yet). My archive.json feed now contains 12 fragments like this:

"_citations": [
  {
    "url": "https://doi.org/10.1038/nchem.2187",
    "doi": "10.1038/nchem.2187",
    "cito": [
      "citesForInformation"
    ]
  }
]

It matches the structure of the _references extension.

Let’s see where this goes :)

References

I have many books in my house on these aspects of science. From text books, to primary literature, to links to blog posts. It is not my research field, and (scientific) communication is very much a research field. “It makes sense”, does not do, but the awareness of some of the things should be basic academic training. Or probably earlier. Every writing has context. Not a random pointer, but a friend has recently written with several others an interesting chapter on reading scientific literature (doi:10.1515/9783110782844-010, pdf).

Scientias

And while I started the Scientias (with (the) Diederik Jekel and Krijn Soeteman) episode to hear more about the science behind “history”, there were several topics in the [Over oudheidkunde, nepvondsten en AI: wat weten we écht over het verleden?] episode about the above things, with Jona Lendering as guest.

The things I heard that I got excited about (enough to sit down and write down these thougths) were small comments, other things were more prominent. One more obvious discussion was that a lot of science news is focusing on new facts, and not the science behind it. Not just the methods, but also the stack of assumptions, how they were combined, valued, etc. As Lendering explains how a researcher does that (translated into English by me, with assumptions on the true intention):

This is a hypothesis. Now I stack this hypothesis on the previous hypothesis, and I stack another hypothesis on that, and then I stack a hypothesis on that.

Most science news outlets find that too complicated (for their audience). To me it feels they think their audience is too stupid to understand the full story. Now, the Scientias Podcast made it a unique selling point to explain how those facts were derived from experiments, because they find the full background and experiments behind our knowledge important, exciting. and relevant. I do.

And if I understood Lendering correctly, he too says there actually is plenty of room of that. The full story is more complicated and I cannot do this justice. There is a growing body of scientific literature that cannot be captured in a single blog post. I will not even try. It will likely just backfire.

And I agree. That is not relevant to them nor their point, but it does explain I was excited about the post. It is always nice to hear like-minded people. I do not hear this position about scientific communication a lot, but very, very much agree and try to practice.

Stacking

Lendering about the book he described in the above quote:

I think the conclusions are absolutely wrong, because it stacked too many uncertainties in a row.

I love this quote. Lendering is talking here about archeology, but it applies to at least chemistry and the life sciences too. I once wrote a grant application to the Swedisch national research funder VR, their equivalent of the NWO on that topic (which got rejected, because the candidate has too many international collaborations, despite scores in the fundable range).

It also very much resonates with the whole topic of LLMs, where exactly that is happening: the stacking of many uncertainties. The real problem of AI since the 2000s is not the availability of great machine learning methods, but the facts to learn from, and how well those methods can put those facts in perspective (think Red Riding Hood). That is why my research focus swifted from chemometrics (the old name for AI in chemistry) and knowledge representation, to knowledge representation and interoperability.

The science behind archeology and history

Most of the podcast is about archeology and how history is discovered. And the whole podcast does a wonderful job at explaining how much has changed in the past 30 years. What I learned as a kid about the Romans has been updated in many ways. The the oldest known documented genocide, in Belgica that started with a Roman military camp on the hill that I can see from my work office, has peaked my interest in Roman history. But what fascinates me most are not the narratives but the science behind it. The new chemical and biological approaches giving new evidence that puts archeological evidence in a richer context. Just listen (or watch) the podcast. And I haven’t read it, but I understood Oudheidkunde is een wetenschap covers a good bit of it.

Do I think that Nijmegen is the oldest Dutch city, and not Maastricht? Of course, I lived in Nijmegen for a large part of my life. The competition between the place I studied before and study now is just great amusement. Bring on the popcorn. And that aquaduct in Nijmegen? I cycled everyday on a Roman dike used for that aquaduct. As a researcher, I really do not care the exact dates, or what material that aquaduct was made of. How we could learn those things, well, yeah, the chemistry of physics behind that is cool too. Bring on the nerdy science.

What science should be like

Another topic in the above theme is what science should be. Lendering is negative about the research schools and research institutes (in the Dutch implementation) and explains that different disciplines should communicate and collaborate more. Since my field is in between two established research fields, I can painfully confirm that many (Dutch) researchers are much to focused on their own narrow specialism, that they lost context to put that specialism in a wider context.

You only start writing a publication, after you collected and looked at all evidence.

Lendering then talks about how modern science may have deviated too much from this. For people who have been following the Open Science news, they have seen a good body of scientific literature that show that this indeed is not always to case (sic). For example, Lendering states:

There is no substantial peer review anymore. Really weird publications happen, that should never have been published.

Really, bold, perhaps, but sadly too close to reality with enough primary literature that has studied these issues (some of which you can find convered in my blog posts). Commercial publishers expect a peer review in 10 working days. They value speed over quality. It is there business model. But also researchers that simply submit the manuscript as-is to the next journal, when the previous journal clearly outlined limitations. There are reasons why I had to step down as editor from Springer Nature.

Calls for Action

Before I write a blog post that takes you more time than to listen to the podcast, there are two more quotes I like to cover:

I could not resist the opportunity to mention paywalls. And what has been behind paywalls is not released as open access. And there are not plans for this.

Lendering continues with a good example from desinformation on why this must change. Again, I am very excited about this statement, which I have been arguing for too.

The next quote I shortened a bit, because I want you to listen to the full discussion in the podcast, but here I do not want to focus on the very convincingly bad urgency of the given example:

there is a theory from the 19th century, and that theory has been disproved since then. [..] But due to digitization projects, old literature is available, with the old disproven theories too. This is in itself not an issue. However, because the correcting literature from the last 30 years are hidden behind the paywall, is the disproven theory unchallenged.

There are many aspects of just this part, and many relate to how knowledge and how we got to knowledge is spread, reused, etc. Bascially, the FAIR data principles are in that sense just a reformalisation of a much older and bigger problem. And I think Lendering is spot on with this observation. Translating this to the bigger issues, the commercial publishers can lecture researchers about FAIR data, but as long as the with strong determination frustrate scientific progress with paywalls, etc, then we will not make the progress we deserve and need.

Action 1: all Dutch literature must become green Open Access

While The Netherlands is just a small player, our national effort towards Open Access has made a clear world wide impact. So has Open Science, and FAIR data, two other (but distinct) movements to improve knowledge dissemination, and both where The Netherlands has made a significant international contribution.

So, Dutch universities should work harder to make the next step, after CC-BY licensing and the Taverne green Open Access law. Let’s remove that small print of 100% Open Access that says, “but only of new literature”. Open scientists have been working on this for long enough; it is just policy, laws, and the mere willingness. Practically, every Dutch university can just tomorrow contact all (emeritus) professors and make all their literature available under the Taverne rules. Many university libraries have already started this, but this must have UNL backing with a clear mandate to make 100% of all Dutch works available, across the full history.

I repeat, this is not a technical question. It is just doing it.

This goes for every Dutch researcher that want to have their research have the impact it can have. Go to your university library, and send them publisher PDFs of all articles (and book chapters) you ever published at that university. Yes, you can even email the library where you did your PhD.

Taverne is not the full story. Authors have to give permission, via opt-in or opt-out arrangements with their universities. That is a problem when the researcher is no longer around. But can we at least start with all that paywalled literature of which researchers can give that permission, please, for the full appointments, going back in time? For the rest, well, we have to start lobbying with our government. That is quite modern and should be trivial.

I think this is essential: you cannot fully understand the impact of an article if you do not have access to the cited and citing literature because it is behind a paywall.

Action 2: make the context of literature easier to record and easier to access

The other point Lendering made is about the self-correcting nature of science (or lack thereof). The peer review problems were mentioned in the podcast, as well as people not easily finding literature that disproves earlier literature. It can be a lot simpler to get informed about the status of any research work. But right now, even something as simple as getting PubPeer and RetractionWatch information about a References list is non-trivial. The choice of commercial publishers to favor PDF over (scientific) HTML has been one driven by perverse incentive, not by scientific needs. Similarly, it is trivial to learn that an article has been disproven, but not simple. University libraries are not yet equiped correctly or funded appropriately to allow researchers (and the general local audience who can also enter university libraries to learn stuff).

My gauntlet to Springer Nature about CiTO citation intent annotations has yet been underwhelming, despite an in my opinion hopeful and promising pilot. I cannot complain, and the interest is there, but it was too difficult for Springer Nature. And with that we are back with the one of the other main themes of the podcast.

With that, I isolated just a few moments from the podcast. I insist you listen to the full podcast to get the proper context of the above quotes. If people want to learn more about the context I see, I invite you to browse my blog (newer posts and older posts).

Finally, thank you Krijn, thank you Jona, for this awesome 83 minutes and 16 seconds at 1.25x speed!

References

Many things came up. One was testing of new InChI functionality for these organometallic compounds, particularly the stereochemistry. Wikipedia has many chemical compounds and could be a source, but so does Wikidata. Both use the SMILES, but not all SMILES captures all the chemistry we need. And the InChI software needs an V3000 MDL Molfile. Thanks to John and other CDK developers, there is good support for recent cheminformatics software, but I was not sure it had what I would need.

This post is the first of a few related posts. This post is about converting SMILES from Wikidata to V3000 files. Take cisplatin: it has four ligands around a platinum atom, all in a single plane:

In this image, in red, is actually an annotation of how the ligands are oriented around the platinum. This is also reflected in the isomeric SMILES in Wikidata: Cl[Pt@SP1]([NH3])([NH3])Cl.

The following source code is written in Groovy which I have used for many years because it is less verbose than Java. First, we set up our helper classes:

@Grab(group='org.openscience.cdk', module='cdk-smiles', version='2.12')
@Grab(group='org.openscience.cdk', module='cdk-silent', version='2.12')
@Grab(group='org.openscience.cdk', module='cdk-ctab', version='2.12')
@Grab(group='org.openscience.cdk', module='cdk-sdg', version='2.12')

import org.openscience.cdk.smiles.SmilesParser;
import org.openscience.cdk.silent.SilentChemObjectBuilder;
import org.openscience.cdk.io.SDFWriter;
import org.openscience.cdk.layout.StructureDiagramGenerator;
import javax.vecmath.Vector2d

builder = SilentChemObjectBuilder.getInstance()
sp = new SmilesParser(builder)
sdg = new StructureDiagramGenerator();

With some extra code, I can actually get many compounds from Wikidata to convert to v3000 with a SAPRQL (as I have done last year with CXSMILES and polymers, unpublished), but let’s go with a single example:

smiles = "Cl[Pt@SP1]([NH3])([NH3])Cl"
label = "cisplatin"
wdItem = "Q412415"

I can parse the SMILES and generated 2D coordinates with (which is also the approach by CDK Depict which I used for the above 2D diagram of cisplatin):

mol = sp.parseSmiles(smiles)
sdg.setMolecule(mol);
sdg.generateCoordinates(new Vector2d(0, 1));
mol = sdg.getMolecule();

If you have more than one molfile, they can be combined into a SD file, to which additional properties can be added:

mol.setTitle(label)
mol.setProperty("PUBCHEM_SUBSTANCE_SYNONYM", label)
mol.setProperty("PUBCHEM_SUBSTANCE_COMMENT", smiles)
mol.setProperty("PUBCHEM_EXT_DATASOURCE_REGID", wdItem)
mol.setProperty("PUBCHEM_EXT_SUBSTANCE_URL", "https://qlever.scholia.wiki/" + wdItem)

And then generate the actual SD file with:

writer = new FileWriter(new File("demo.sdf"))
SDFWriter sdfWriter = new SDFWriter(writer);
sdfWriter.getSetting(SDFWriter.OptAlwaysV3000).setSetting("true");
sdfWriter.write(mol);
sdfWriter.close();
writer.close();

We then get this v3000 file:

cisplatin
  CDK     07262618042D

  0  0  0     0  0            999 V3000
M  V30 BEGIN CTAB
M  V30 COUNTS 5 4 0 0 0
M  V30 BEGIN ATOM
M  V30 1 Cl -1.29904 2.25 0 0
M  V30 2 Pt 0 1.5 0 0
M  V30 3 N 1.29904 2.25 0 0 VAL=4
M  V30 4 N 1.29904 0.75 0 0 VAL=4
M  V30 5 Cl -1.29904 0.75 0 0
M  V30 END ATOM
M  V30 BEGIN BOND
M  V30 1 1 2 1 CFG=3
M  V30 2 1 2 3 CFG=3
M  V30 3 1 2 4 CFG=1
M  V30 4 1 2 5 CFG=1
M  V30 END BOND
M  V30 END CTAB
M  END
> <PUBCHEM_SUBSTANCE_COMMENT>
Cl[Pt@SP1]([NH3])([NH3])Cl

> <PUBCHEM_EXT_DATASOURCE_REGID>
Q412415

> <PUBCHEM_SUBSTANCE_SYNONYM>
cisplatin

> <PUBCHEM_EXT_SUBSTANCE_URL>
https://qlever.scholia.wiki/Q412415

$$$$

I can copy/paste the resulting v3000 content to the InChI Web Demo to calculate the Standard InChI:

InChI=1S/2ClH.2H3N.Pt/h2*1H;2*1H3;/q;;;;+2/p-2

And this is what the two technical meetings I attended were about: molecular inorganics. The above InChI does not feel right, and certainly lost the connectivity of the ligands with the platinum. However, if we add the beta option -MolecularInorganics, then we get this InChI (where the B in InChI=1B reflects the beta state of this feature):

InChI=1B/Cl2H6N2Pt/c1-5(2,3)4/h3-4H3

However, this beta version does not distinguish cisplatin from transplatin. For that, we need to dive into how to represent the stereochemistry of these inorganics first.

One aspect that limits the uptake of diamond open access models is that the scholarly community needs to change their trust model. We have seen a transition from societal publishers to commercial publishers, from scholarly-led to publisher-led. And commercial publishers certainly breached my trust (see also doi:10.5281/zenodo.4926030).

Many scholars prefer the certainty of a scholarly journal where they can trust that the editorial board and their reviewers take their job seriously. This matters. The notion that peer reviewers may reject your work is an extrinsic motivation for at least some researchers to do a better job (personal communication). And that feature works well: there are indications that research published as preprint sees limited change after the formal peer review process (doi:10.1038/d41586-026-02167-3 and doi:10.64898/2026.06.30.735556):

“For most scientists, reputation is important, and they would not upload something they would not be comfortable with publishing in a journal later.”

In an ideal situation, reputation would not be part of the whole equation: researchers never submit anything they think is not scientifically sound. Journals would never desk-reject an article that is scientifically sound. The question is mostly how to reach that. Everyone is busy, and it has not been so long ago that the journal impact factor was by many as indicator of scientific quality. The reality is far more complex than that.

Finding Diamond Open Access journals

So, let’s assume you are interested in the idea, and want to know what diamond open access journals exist in your research field, before you can even consider the quality of that journal, you need to know which journal is diamond open access. Therefore, I asked yesterday on Mastodon about databases that list which journal are diamond.

It actually turns out that the Directory of Open Access Journals (DOAJ) does not indicate if a journal is diamond. You can filter on zero APC, but that is not the same. In fact, zero APC is routinely used by big publishers to launch new journals. One they have critical mass, they will start charging non-zero APCs.

I already received some replies about possible options:

1. Diamond Discovery Hub: lists just over 4,000 diamond journals, but seems closed data (thx Najko! see also doi:10.15291/libellarium.4569)
2. wheretopublish.github.io: open data, accepts additions, but lists only 9 diamond journals (thx Anne!)
3. B!SON: a recommender that depends on DOAJ, so also without clear diamond indication (thx Jon!)

All pieces of the puzzle.

When I grew up as a scholar, I learned to publish in journals where you respect the research, where you see important research in your field get published. Nowadays, this is changing. In fact, young researchers (tho many older scholars alike) fear to publish in predatory journal by accident (really! [citation_pending]). The big publishers and the new publishers have been adding so many new journals, and academia has really pushed the “publish as many articles as you can” model, that it kind of makes sense why things are different now.

And, a central database with trustworthy information about the quality of diamond open access journals is just needed.

A final note, while I have not been able to create a list of diamond open access journals with OpenAlex, already in February 2025 I found a way to list articles tagged with a diamond status. This query lists over three thousand of such annoated articles for Maastricht University:

(Interestingly, the journal where the three most cited articles were published turned diamond already in 2004 and was discontinued about half a year go “Due to recent changes in operational resources”, according to Wikipedia.)

Building trust

So, while the above solution do allow you to increase your success rate or finding a diamond open access journal in your field, the next question is: what can you learn about this journal. For example, what does Wikipedia say about the journal (if anything)? Do you know people that have published in that journal? The OpenAlex screenshot shows

That make me wonder what Wikidata has to offer. After all, as one of the largest linked open data sources, they must have some information.

I am not disappointed. The type diamond open-access journal is actively used, and there is plenty of other of useful fields for these journals. I came up with this SPARQL query (with some QLever sugar for labels):

PREFIX wdt: <http://www.wikidata.org/prop/direct/>
PREFIX wd: <http://www.wikidata.org/entity/>
PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#>
SELECT DISTINCT ?diamondJournal ?diamondjournalLabel
      # (SAMPLE(?url_) AS ?website)
      (SAMPLE(?issnl_) AS ?issnl)
      (GROUP_CONCAT(DISTINCT STR(?issn_)) AS ?issn)
      (SAMPLE(?lang_) AS ?language)
      (SAMPLE(?doaj_) AS ?doaj)
      (GROUP_CONCAT(DISTINCT STR(?subject);separator=", ") AS ?subjects)
WHERE {
  ?diamondJournal wdt:P31 wd:Q108440863 .
  OPTIONAL { ?diamondJournal wdt:P236 ?issn_ }
  OPTIONAL { ?diamondJournal wdt:P7363 ?issnl_ }
  OPTIONAL { ?diamondJournal wdt:P856 ?url_ }
  OPTIONAL { ?diamondJournal wdt:P407 / @en@rdfs:label ?lang_ }
  OPTIONAL { ?diamondJournal wdt:P5115 ?doaj_ }
  OPTIONAL { ?diamondJournal wdt:P921 / @en@rdfs:label ?subject }
  ?diamondJournal @en@rdfs:label ?diamondjournalLabel .
} GROUP BY ?diamondJournal ?diamondjournalLabel

This lists more than 1,500 diamond open access journals:

With Wikidata we can then list UM authors that published in diamond journals. That list is remarkably short, however, if you compare it with the earlier results from OpenAlex. That is because Wikidata only contains a subset of scientific literature and not all articles are correctly linked to the items about their authors. But I also spotted journals in the OpenAlex list that are not really diamond.

What is next?

Clearly, we need better metadata, but things are not looking that bad either! We have something to work with. What is your favorite diamond open access journal?

References