CFSME ATLAS site

SNT Gatchaman said:
I imagine that adding tags is doable between database, application layer and user interface, and that the hard part would be adding sensible tags to all these papers. Potentially that could also be crowd-sourced, but I wouldn't be surprised if AI could make a decent first pass.
Click to expand...
I doubt that AI could do this, what I have seen is mostly people using existing PubMed tags or things mentioned in the abstract.

With a team it should be doable to screen all ME/CFS papers on PubMed: there are approximately 7700 at the moment. Many are short commentaries or letters. I (and many others on the forum here) have probably read most of them already.

Valuable things to extract are:
  • The type of publication (letter/commentary, review, experiment, prospective study, randomized trial, etc)
  • The number of ME/CFS participants in the trial or experiment
  • The area of research (genetics, metabolomics, immune system, exercise testing, Chinese medicine etc.)
  • The methods used (cytokine measurement, RNA-seq, CPET, which questionnaires were used, etc)
Things like publication date, journal and authors can be extracted automatically by AI.

One problem is that it probably needs to be updated for new papers that come out, which would make it an ongoing effort that never stops.
 
I think an automated tagging solution is viable. In fact a lot of similar work & potentially useful component elements have already been published, some of which could be adapted (e.g. PubTator 3; BioBERT / PubMedBERT). There is MeSH and the Human Phenotype Ontology. Some kind of hybrid ontology-based extraction and a multi-level classifier might be the way, although I'm sure there are others on the forum who know far more about AI than I do.

One other point I'd make is that many ME/CFS papers are in journals not indexed by MEDLINE and therefore the abstracts will not be retrievable via the PubMed API/Entrez. Anyone building such a tool would probably have to include support for e.g. the EuropePMC REST API for broader coverage.
 
Last edited:
ME/CFS Science Blog said:
With a team it should be doable to screen all ME/CFS papers on PubMed: there are approximately 7700 at the moment. Many are short commentaries or letters. I (and many others on the forum here) have probably read most of them already.
Click to expand...

I made a library system (macOS) and I currently have 4734 ME/CFS papers downloaded and tagged, and most have the relevant S4ME thread URLs. The papers' metadata are included: journal, authors, pub date, PubMedID, abstract etc. I could provide Robius with a JSON file so you could loop through and tag by DOI fairly easily. Obviously that's using my ideas about tagging, which I've sometimes abbreviated, but it could be a useful start. You might want to lowercase and replace spaces with hyphens for example.

Eg Multi-omics identifies lipid accumulation in Myalgic Encephalomyelitis/Chronic Fatigue Syndrome cell lines: a case-control study (2026) looks like (abbreviated, and probably doesn't need the additional escape characters) —

JSON: 
    {
      "authors" : "Missailidis, Daniel; Armstrong, Christopher W; Anderson, Dovile; Allan, Claire Y; Sanislav, Oana; Smith, Paige K; Esmaili, Tammy; Creek, Darren J; Annesley, Sarah J; Fisher, Paul R",
      "dateAdded" : "2026-01-08T23:54:07Z",
      "journal" : "Journal of Translational Medicine",
      "pdfURL" : "https:\/\/link.springer.com\/content\/pdf\/10.1186\/s12967-025-07620-x.pdf",
      "publicationDate" : "2026-01-07T11:00:00Z",
      "read" : true,
      "abstract" : "BACKGROUND\nIn recent years, evidence has indicated a metabolic shift towards increased demand for lipids in various lymphoid cell populations from people with Myalgic Encephalomyelitis\/Chronic Fatigue Syndrome (ME\/CFS). We previously screened  ETC...",
      "doi" : "10.1186\/s12967-025-07620-x",
      "lastOpened" : "2026-04-10T09:19:45Z",
      "notes" : "[S4ME](https:\/\/www.s4me.info\/threads\/multi-omics-identifies-lipid-accumulation-in-myalgic-encephalomyelitis-chronic-fatigue-syndrome-cell-lines-a-case-control-study-2026-missailidis-et.48171\/)",
      "openAccess" : true,
      "pubMedID" : "41508032",
      "rating" : 5,
      "readDate" : "2026-01-09T05:51:39Z"
      "tags" : [
        "Steroids",
        "PTDSS1",
        "Key Papers",
        "Lipid Rafts",
        "ME\/CFS",
        "Genetics \/ GWAS",
        "mTOR",
        "B Cells",
        "Ceramide",
        "Mitochondria",
        "Phosphatidylserine",
        "Glutamate",
        "Lipid Droplets",
        "NLRP3",
        "Glutamine",
        "Sphingolipids",
        "ER",
        "Lipids",
        "Cholesterol",
        "Plasmalogen",
        "Vit B",
        "Bile Acid",
        "Metabolism",
        "CD36"
      ],
      "title" : "Multi-omics identifies lipid accumulation in Myalgic Encephalomyelitis\/Chronic Fatigue Syndrome cell lines: a case-control study",
      "url" : "https:\/\/link.springer.com\/article\/10.1186\/s12967-025-07620-x"
    },
 
SNT Gatchaman said:
I made a library system (macOS) and I currently have 4734 ME/CFS papers downloaded and tagged, and most have the relevant S4ME thread URLs. The papers' metadata are included: journal, authors, pub date, PubMedID, abstract etc. I could provide Robius with a JSON file so you could loop through and tag by DOI fairly easily. Obviously that's using my ideas about tagging, which I've sometimes abbreviated, but it could be a useful start. You might want to lowercase and replace spaces with hyphens for example.

Eg Multi-omics identifies lipid accumulation in Myalgic Encephalomyelitis/Chronic Fatigue Syndrome cell lines: a case-control study (2026) looks like (abbreviated, and probably doesn't need the additional escape characters) —

JSON: 
    {
      "authors" : "Missailidis, Daniel; Armstrong, Christopher W; Anderson, Dovile; Allan, Claire Y; Sanislav, Oana; Smith, Paige K; Esmaili, Tammy; Creek, Darren J; Annesley, Sarah J; Fisher, Paul R",
      "dateAdded" : "2026-01-08T23:54:07Z",
      "journal" : "Journal of Translational Medicine",
      "pdfURL" : "https:\/\/link.springer.com\/content\/pdf\/10.1186\/s12967-025-07620-x.pdf",
      "publicationDate" : "2026-01-07T11:00:00Z",
      "read" : true,
      "abstract" : "BACKGROUND\nIn recent years, evidence has indicated a metabolic shift towards increased demand for lipids in various lymphoid cell populations from people with Myalgic Encephalomyelitis\/Chronic Fatigue Syndrome (ME\/CFS). We previously screened  ETC...",
      "doi" : "10.1186\/s12967-025-07620-x",
      "lastOpened" : "2026-04-10T09:19:45Z",
      "notes" : "[S4ME](https:\/\/www.s4me.info\/threads\/multi-omics-identifies-lipid-accumulation-in-myalgic-encephalomyelitis-chronic-fatigue-syndrome-cell-lines-a-case-control-study-2026-missailidis-et.48171\/)",
      "openAccess" : true,
      "pubMedID" : "41508032",
      "rating" : 5,
      "readDate" : "2026-01-09T05:51:39Z"
      "tags" : [
        "Steroids",
        "PTDSS1",
        "Key Papers",
        "Lipid Rafts",
        "ME\/CFS",
        "Genetics \/ GWAS",
        "mTOR",
        "B Cells",
        "Ceramide",
        "Mitochondria",
        "Phosphatidylserine",
        "Glutamate",
        "Lipid Droplets",
        "NLRP3",
        "Glutamine",
        "Sphingolipids",
        "ER",
        "Lipids",
        "Cholesterol",
        "Plasmalogen",
        "Vit B",
        "Bile Acid",
        "Metabolism",
        "CD36"
      ],
      "title" : "Multi-omics identifies lipid accumulation in Myalgic Encephalomyelitis\/Chronic Fatigue Syndrome cell lines: a case-control study",
      "url" : "https:\/\/link.springer.com\/article\/10.1186\/s12967-025-07620-x"
    },
Click to expand...
Thanks, I really appreciate the offer. It could definitely be useful.
At the moment though I think I need to focus on stabilising, organising, and developing what I already have first. I’m also still not finished with importing, processing, and analysing the studies already sitting in the pipeline / backlog, so I probably shouldn’t take on another large input just yet.
But I’m very grateful for the offer, and I may come back to it a bit later when the current system is in a more settled state.
 
Have you thought of structuring the website maybe more like a wiki. In the sense of letting people fairly simply contribute, for example, tags if they want to.

I realise this is not something you can implement very easily but it seems like having an easy pipeline for others to contribute and discuss is something that would be helpful to keep the project sustainable.

Just something like having an input where people can suggest edits or tags could help IMO.
 
Yann04 said:
Have you thought of structuring the website maybe more like a wiki. In the sense of letting people fairly simply contribute, for example, tags if they want to.

I realise this is not something you can implement very easily but it seems like having an easy pipeline for others to contribute and discuss is something that would be helpful to keep the project sustainable.

Just something like having an input where people can suggest edits or tags could help IMO.
Click to expand...
Yes, I want to add more layers of public interaction in different ways later on, but it needs to be very carefully thought out and precisely done.
 
ME/CFS Science Blog said:
I doubt that AI could do this, what I have seen is mostly people using existing PubMed tags or things mentioned in the abstract.

With a team it should be doable to screen all ME/CFS papers on PubMed: there are approximately 7700 at the moment. Many are short commentaries or letters. I (and many others on the forum here) have probably read most of them already.

Valuable things to extract are:
  • The type of publication (letter/commentary, review, experiment, prospective study, randomized trial, etc)
  • The number of ME/CFS participants in the trial or experiment
  • The area of research (genetics, metabolomics, immune system, exercise testing, Chinese medicine etc.)
  • The methods used (cytokine measurement, RNA-seq, CPET, which questionnaires were used, etc)
Things like publication date, journal and authors can be extracted automatically by AI.

One problem is that it probably needs to be updated for new papers that come out, which would make it an ongoing effort that never stops.
Click to expand...
I’d be up for a dull admin task which requires very low skills like categorising papers by type and number of participants (I am famously not at all scientific).
 
Just to clarify a bit, for now new research only comes in from PubMed, but that part is already built and working automatic. New papers get fetched, then goes through classification and categorisation before it gets integrated across the Atlas, so it is not only raw PubMed tags or things from the abstract put straight on the site. There is also an automatic weekly email every Monday, and I got the latest one yesterday, so that seems to be working fine too. Later on it can be expanded beyond PubMed as well. I also have many other ideas I want to work on over time, but I need to take it a bit slow because my health is a bit down at the moment. Slow is fast. I really appreciate the thoughts here, and I am very open to ideas, criticism and suggestions. Any help or input is genuinely welcome.
 
Hi there Robius and others interested in Robius's project,

In the recent past, I've thought about working on initiatives that try to bridge the gap between MECFS scientific literature and those downstream of it (patients, advocates, but also doctors and other researchers). I came up with some challenges for my own work that I believe can serve as feedback for your project.

Challenges
1. Where does authority come from? How to establish trustworthiness?

Some thoughts:
  • On the use of AI: for me, reading even just a few AI-generated sentences breaks my trust in a source. There may still be great value in what's written, but it requires a lot of work from the reader to validate. A vibe-coded UI reduces my trust in your source even further.
    This is coming from someone who uses LLMs daily for software and pays a monthly subscription.
  • I generally place more trust in authored content than nameless publications.
  • Wikis like MEPedia are great, but being editable by anyone does affect their authority.
  • I believe the trustworthiness of this forum comes from its open, critical format and scientific values, but also in large part from the credentialed researchers who contribute to it. (Still, it may raise eyebrows when I mention "something I read on an internet forum").
2. Don't become just another source on MECFS, diluting the information landscape. Avoid duplicating effort with similar initiatives.

3. Find a balance between scientific neutrality and bringing a clear message against harmful practices and wasteful research.

4. Challenges related to judging papers:
  • Hostility from authors
  • Maintain objectivity
  • Maintain accuracy
  • Who does the work? Judging trial setup may be easier than evaluating biomedical or technical aspects.
5. Challenges around collaboration, moderation, and sustainability.

I found this thread about MEPedia that touches on some of these challenges.

Challenges aside, I think your initiative is good and it's nice to see there's some enthusiasm. As I wrote in my welcome post, I'd like to contribute to making the forum's ideas more accessible.

I'll continue my thoughts in another comment when I have more time.
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom