Review Machine learning and multi-omics in precision medicine for ME/CFS, 2025, Huang....Armstrong

[Jonathan Edwards]

I am sure the Melbourne team have the best of intentions here but I do agree that this is not a very well written review.

I used to insist that doctoral students wrote a review before I offered them a doctoral attachment, to see if they could think critically enough for it to be worth them going for a doctorate, but I rarely suggested that they published the result, which by definition was not an expert view. The literature is groaning under the weight of reviews written by novices.

I am currently supervising a bachelors literature review project and this reminds me rather of my student's first draft. I scrawled at the top: Don't write as if you are an expert narrating history and words of wisdom from the literature. Write as if you don't necessarily believe a bloody word of it and are very dubious about loads of assumptions people have made.'

To me 'precision medicine' is a phoney buzzword. One day such a thing might emerge but inasmuch as personalised medicine exists it has probably been there all along and is much less a feature than it was fifty years ago.

Omics technology is progressing and is fascinating, but why not have at least one interesting example of a new development in the abstract? Without it I didn't read on I am afraid.

Sorry to be critical, but that's the name of the real science game.

And surely defying a simplistic characterisation is a platitude. Simplistic means 'making a description too simple'. (It doesn't just mean simple.) Everything defies characterising by a too simple characterisation!

 

[Hutan]

ME/CFS presents new challenges to traditional healthcare
The traditional procedure for classification of a disease proceeds by identifying the primary dysfunctional organ in which the cardinal symptoms manifest [18]. ME/CFS does not fit neatly into this approach as symptoms can arise from the musculoskeletal, immunological, cardiovascular, gastrointestinal, and neuroendocrine systems (Fig. 1).

Reference #18 is A Treatise on the Principles and Practice of Medicine. Designed for the Use of practitioners and students of Medicine. Atlanta Med Surg J. 1867!
I don't think it is helpful suggesting that ME/CFS is such an exception from other diseases.

Think of sickle cell anaemia - it has symptoms of pain throughout the body, fatigue, infections, vision problems, diarrhoea. Prior to 1910 when it was first documented in western medicine, many people with the condition probably would have been in the same position as people with ME/CFS are in now in terms of there being little understanding and no treatment. There are lots of diseases that have symptoms in a range of body systems.



The idea of Figure 1 is really good. It summarises 'symptom manifestations and biological characterisation opportunities'.

A notable gap though is muscle fatiguability, which is a major and common symptom of ME/CFS. The Musculoskeletal heading has only

• muscular/joint pain
• muscle twitching
• hypermobility

I don't think we've seen much evidence yet to support muscle twitching and hypermobility as key symptoms of ME/CFS. Related to muscle fatiguability are a range of biological characterisation opportunities such as gait studies, eye muscle fatiguability studies and muscle electrical conduction studies - but none are listed in the figure.

Under Cardiovascular is

• Orthostatic hypotension
• Irregular heart beat
• Dizziness
• Chest pain

I'm not sure that I have seen irregular heart beat as an evidenced symptom of ME/CFS, or chest pain. I think high resting heart rate is a more commonly discussed symptom. And it's odd to single out orthostatic hypotension for inclusion, when really I think orthostatic intolerance is the symptom, and various sorts of oddities with blood pressure and heart rate are opportunities for biological characterisation.

There are probably more improvements that could be made to the figure e.g. blurred vision.

 

[Peter T]

chest pain

I don’t know how common it is, but chostochondritis is regularly reported in social media.

With me it varies with my ME, and when I was at my worst the pain was one of my most disabling symptoms. Currently it is only there when I am in PEM. Having said that it only began over fifteen years plus into my ME.

It is frustrating that we still lack a comprehensive ‘natural history’ of our condition.


Mod note: We've created a new symptom discussion thread and have moved the costochondritis discussion there.
Costochondritis and rib pain

 

[Hutan]

There are some good points made e.g.

The fluctuation of symptoms and therefore 'omics measurements

This is relevant for ME/CFS, where symptoms and severity fluctuate, and patients experience periods of relative wellness and exacerbation. Thus, there are limitations to relying on single measurements taken at an isolated timepoint, emphasising the importance of capturing dynamic changes over time in the individual, for both research and clinical settings. Once continuous data is collected, machine learning algorithms such as time series forecasting and anomaly detection, can be employed to model an individual’s baseline, identify deviations from this baseline, and predict adverse events, such as PEM accordingly [26].

New technologies such as wearables, home testing, better quicker analysis

The integration of advanced technologies such as wearable devices (for continuous passive data collection), at-home testing kits (for convenience), and high-throughput profiling now enables repeated, real-time measurements and standardised dynamic data collection, which were not previously accessible or widely utilised.

Treatment strategies for ME/CFS prioritise managing symptoms through social, physical, and occupational support, and energy conservation via pacing [24].

I don't think pacing is a treatment strategy and nor is support of various sorts. These are management strategies. They don't cure the disease. Again, it is important that the ME/CFS literature acknowledges that we don't have evidenced treatments. Even things like pain relief are not treatments of ME/CFS, they are treatments of pain or whatever.

However, similar symptoms can arise from different aetiological events and pathophysiological mechanisms, causing varied responses to the same therapies. Although both pharmacologic and nonpharmacologic interventions are available to alleviate symptoms [29], they are often prescribed on a trial-and-error basis. This uncertainty has led to a growing online community of ME/CFS patients sharing their self-medication experiences, highlighting the urgent need for personalised treatment approaches that target the underlying biology of the individual.

Yeah, this is where I get concerned again. It's more of the 'there are treatments, it's just that no one treatment helps everyone - so we just need to work out how to match the treatments to the person better'.

There is not an 'urgent need for personalised treatment approaches that target the underlying biology of the individual' - that is not the first answer to a 'growing online community of ME/CFS patients sharing their self-medication experiences'. The urgent need is finding out what causes ME/CFS and doing quality trials to find a treatment that helps. You don't run before you can walk.

 

[Hutan]

However, similar symptoms can arise from different aetiological events and pathophysiological mechanisms, causing varied responses to the same therapies. Although both pharmacologic and nonpharmacologic interventions are available to alleviate symptoms [29], they are often prescribed on a trial-and-error basis. This uncertainty has led to a growing online community of ME/CFS patients sharing their self-medication experiences, highlighting the urgent need for personalised treatment approaches that target the underlying biology of the individual.

Reference 29 is
Kavyani B, Lidbury BA, Schloeffel R, Fisher PR, Missailidis D, Annesley SJ, et al. Could the kynurenine pathway be the key missing piece of myalgic Encephalomyelitis/Chronic fatigue syndrome (ME/CFS) complex puzzle? Cell Mol Life Sci. 2022;79(8):412.
That was a bad paper. We talked about it here:
Could the kynurenine pathway be the key missing piece of Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) complex puzzle?, 2022, Kavyani
The pharmacological treatment reported for fatigue was stimulants and vitamin and mineral supplements.

In addition to pharmacological interventions, many non-pharmacological therapies have been studied; including psychological methods (i.e., Cognitive behaviour therapy (CBT)) and exercise (such as Graded exercise therapy (GET), rehabilitation and acupuncture) [8, 9]. However, none of the proposed therapies are of proven benefit to all patients

I'm conscious that these papers are being done by students. I'm grateful that they have decided to devote years of their lives to moving the understanding of ME/CFS forward. But, they and their supervisors need to be aware that if they are going to improve things for us, they need to get things right rather than confusing things further. The thing is, the paper isn't just about things like computational biology techniques, it's about a disease affecting people. Those of us with the disease want to see that the people we are relying on to find the answers both understand ME/CFS and care enough to get things right.

I think @Jonathan Edwards has a point when he suggests that not every review needs to be published. Either that or the supervisors need to put more time into improving the student's draft, polishing it for publication. Perhaps getting the student to post their draft here would be one way to get pre-publication feedback. It's annoying for everyone when criticism comes after publication.

 

[Jonathan Edwards]

However, none of the proposed therapies are of proven benefit to all patients

Yes, we should all really put in a plea, however much we respect the research group, for keeping statements like this out. None of the treatments have reliably been shown to be of benefit to anyone. If people are going to review clinical studies they need to have adequate understanding of the basics of interpretation.

 

[Trish]

Perhaps students in this situation could be encouraged to share their learning process and review writing process here, so we can give feedback before it reaches publication stage. They could do it on a members only thread if they don't want to make their learning process public.

We all want research students to do well both for themselves as they get stuck into the process of learning to be good researchers, and for pwME who will benefit from better quality research and publications.

 

[Peter T]

Perhaps students in this situation could be encouraged to share their learning process and review writing process here, so we can give feedback before it reaches publication stage. They could do it on a members only thread if they don't want to make their learning process public.

We all want research students to do well both for themselves as they get stuck into the process of learning to be good researchers, and for pwME who will benefit from better quality research and publications.

I think this would very much depend on the student, we have had students participating here that have coped well with the cut and thrust of the debate but also others who fairly promptly disappear. It does require a certain robustness.

I will never forget tutorials with one tutor from a very academic family going back several hundred years, who would rip my essays metaphorically into tiny shreds and leave me a cowering wreck in the corner, to then conclude ‘a excellent essay Peter, but you do see where you went beyond the evidence’. Once I got used to this, I welcomed it, but I was aware of others that never got used to it, even assuming she was deliberately malevolent.

 

[Yann04]

I think this would very much depend on the student, we have had students participating here that have coped well with the cut and thrust of the debate but also others who fairly promptly disappear. It does require a certain robustness.

Yes if I was a student and my first experience on this forum was getting heavy criticism of something I spent ages working on, I think I’d take it okay at first but if it kept piling on (like it often does here), I’m not sure I’d be able to not take it personally, I’d probably associate this forum with a negative feeling and not come back.

I think it’s great that forum members are able to find so many problems with works, and I think we are especially good at that, but a side effect is that it probably isn’t the most welcoming place for someone who is publishing as a student (I mean positive reinforcement tends to work far better than negative reinforcement to teach people things).

 

[Trish]

That's why I suggested joining the forum before submitting their essay for publication. I think we have been quite helpful to students who ask questions and want to understand more, especially if they tell us what stage they are in learning. The problem comes when journals publish student essays, so we critique them by the same standard as we critique their professors.

I hope the student in this case will treat it as a learning experience, a tough one, but one that can help them become a better scientist.

Take into account also that forum threads are conversations between people with vastly varied backgrounds. Some of us get things wrong, some have illuminating insights. Part of learning is acquiring the knowledge and skills to distinguish between them.

 

[Utsikt]

I've always had a bit of a problem with 'poorly understood' which is trotted out in countless preambles and abstracts. It implies that some actual understanding of the condition exists, which doesn't seem to be the case. Why not just say that it isn't understood?

What isn’t understood?

If someone meets the criteria for ME/CFS after a Covid-19 infection, they have ME/CFS. Someone with ME/CFS after a Covid-19 infection doesn't really have 'long Covid' as a co-morbidity, anymore than someone with ME/CFS after EBV has 'long EBV' as a co-morbidity.

Why does ME/CFS exclude LC as a co-morbidity? I’ve always thought of LC and ME/CFS as two partially overlapping circles in a Venn-diagram.

I am currently supervising a bachelors literature review project and this reminds me rather of my student's first draft. I scrawled at the top: Don't write as if you are an expert narrating history and words of wisdom from the literature. Write as if you don't necessarily believe a bloody word of it and are very dubious about loads of assumptions people have made.'

This.

Perhaps students in this situation could be encouraged to share their learning process and review writing process here, so we can give feedback before it reaches publication stage. They could do it on a members only thread if they don't want to make their learning process public.

Excellent idea! Although members only threads are quite public given the size of this field. However, I can’t think of a better way to learn, and completely for free as well from people that want to make you the best researcher that you can be.

I think this would very much depend on the student, we have had students participating here that have coped well with the cut and thrust of the debate but also others who fairly promptly disappear. It does require a certain robustness.

What’s the rate of robustness among the BPS gang? I want researchers dedicated to knowledge and science, no matter what it means for them personally. If you’re unable to face being told that you’re wrong, science isn’t for you. The bluntness doesn’t help, but SoMe is far worse.

 

[Utsikt]

Yes if I was a student and my first experience on this forum was getting heavy criticism of something I spent ages working on, I think I’d take it okay at first but if it kept piling on (like it often does here), I’m not sure I’d be able to not take it personally, I’d probably associate this forum with a negative feeling and not come back.

I understand your concern, but I’m not sure I want ME/CFS-researchers that takes it personally when they are proven wrong. That’s partially why we’re in this BPS mess in the first place. We need more quality, not more «anyone’s best».

Regardless, ME/CFS isn’t intuitive. It doesn’t fit most people’s internal model of how a body works. They would have to recalibrate that model sooner or later anyways if they want to contribute anything of value.

It’s not like I could hide from PEM when my entire life was turned upside down and I couldn’t do anything right. Might be good to experience some of that and learn to deal with it. They’ll benefit from it for the rest of their lives.

 

[Yann04]

I understand your concern, but I’m not sure I want ME/CFS-researchers that takes it personally when they are proven wrong.

Yes but most of what this forum touches on when talking about works by students is criticism of terminology, methodology etc

Things where there isn’t really a clean category of right or wrong.

(I’m not denying that there some terrible pseudoscientific stuff going on, but that’s not published by the average student, nor is that present on this paper).

 

[wigglethemouse]

I don’t know how common it is, but chostochondritis is regularly reported in social media.

Just wanted to say I agree with what you say about the chest pain / chostochondritis. I get it too as you describe. Thank you for mentioning it. It's definitely not just one leg longer than the other which is just a silly statement. I liked @Hutan's suggestion that their may be a postural component from lying down in an awkward postion.

Yes if I was a student and my first experience on this forum was getting heavy criticism of something I spent ages working on, I think I’d take it okay at first but if it kept piling on (like it often does here), I’m not sure I’d be able to not take it personally, I’d probably associate this forum with a negative feeling and not come back.

I agree that this thread feels like piling it on over a trivial first paper for a phd student.

 

[Utsikt]

Yes but most of what this forum touches on when talking about works by students is criticism of terminology, methodology etc

Things where there isn’t really a clean category of right or wrong.

There are better and worse ways of doing things, and there’s a threshold along that scale that says «good enough for science». If you make a conscious or unconscious decision to say something over something else, and the chosen way isn’t «good enough for science», then you’re «wrong» in all practical senses of the word «wrong».

These things matter and I’d argue that it’s better to have them ironed out sooner rather than later. The professor should probably warn the students though, and encourage them to use it for learning. Test yourself and become better because of it.

 

[Jonathan Edwards]

I agree that this thread feels like piling it on over a trivial first paper for a phd student.

But surely it is all pretty polite and good humoured and these things ought to be said. Otherwise the whole of medical science will just sink to the bottom of a sea of uncritical words.

 

[Hutan]

Why does ME/CFS exclude LC as a co-morbidity? I’ve always thought of LC and ME/CFS as two partially overlapping circles in a Venn-diagram.

There are two different things - the assumed trigger and the collection of symptoms.

People can meet ME/CFS criteria after a range of triggers, e.g. post-Q fever ME/CFS; post-EBV ME/CFS; post-SARS-1 ME/CFS; Post-COVID-19 ME/CFS. It's often useful to note the trigger, but ME/CFS is the syndrome.

Long Covid is simply the persistence of (new) symptoms for 3 months after Covid-19. For some people, that is loss of taste or smell. That obviously isn't ME/CFS. Neither is lung damage with persistent cough or kidney damage. Long Covid is the umbrella term for a range of health conditions thought to have been triggered by Covid-19.

Some people use Long Covid as a synonym for ME/CFS-like illness after Covid-19 (where not everyone exactly meets ME/CFS criteria, but fatigue is a core part of their illness), but I think that's too imprecise to be useful in science when others are using it to also mean all those other health consequences. If it is used in that way, it needs to be carefully defined, especially in scientific papers.

Yes, you can have ME/CFS and also have a persisting loss of taste or a persisting cough, for example, after Covid-19. But that isn't ME/CFS with a comorbidity of Long covid, it's ME/CFS with a comorbidity of e.g. ageusia or scarred lungs.

Some people who had ME/CFS get more severe after Covid-19. They will probably want to claim having Long Covid, as there can be more help for people with Long Covid than with ME/CFS. But, there is no co-morbidity there.

Many people who develop ME/CFS after January 2020 are likely to attribute it to Covid-19 regardless of the actual trigger and call it Long Covid, because there is more information and awareness out there about Long Covid for doctors and patients and there is more help for people with Long Covid. If that keeps happening, very few people will be labelled with new onset ME/CFS, maybe just some people who develop ME/CFS after another obvious infection such as EBV.

I think it is important that scientists bring clarity to the question of Long Covid. I think that usually means using other terms that are more accurate, or at least explaining what they mean when they use the term.

 

[Hutan]

I think @Jonathan Edwards has a point when he suggests that not every review needs to be published. Either that or the supervisors need to put more time into improving the student's draft, polishing it for publication. Perhaps getting the student to post their draft here would be one way to get pre-publication feedback. It's annoying for everyone when criticism comes after publication.

Perhaps students in this situation could be encouraged to share their learning process and review writing process here, so we can give feedback before it reaches publication stage. They could do it on a members only thread if they don't want to make their learning process public.

I think it’s great that forum members are able to find so many problems with works, and I think we are especially good at that, but a side effect is that it probably isn’t the most welcoming place for someone who is publishing as a student (I mean positive reinforcement tends to work far better than negative reinforcement to teach people things).

A number of us have recognised that it is hard on students to be subjected to public criticism and have tried with this team and others to establish ways for informed patients to provide private feedback before a paper is published. I personally spend time providing private feedback on studies for other teams. It is stressful to make these criticisms here; I worry that I'm not being considerate of the feelings of the PhD student. But, this isn't an essay submitted for a university class, it is a published paper.

I agree that this thread feels like piling it on over a trivial first paper for a phd student.

Regardless of whether the paper is produced by a PhD student or not, published papers have consequences, it's hard to know ahead of time whether they will be trivial or not. We've seen that with the poor quality Kavyani paper, which is cited by this paper and which has contributed to the idea that there are useful treatments for people with ME/CFS, it's just that not everyone responds to all of them and we just don't know who responds to which one. That idea, and probably that paper, has contributed to a situation where people with ME/CFS, often encouraged by their doctors, spend a lot of money trying all sorts of treatments, some of them risky (edit - sometime even though preliminary studies showed no real benefit). The message has been picked up here and amplified further, with the suggestion that precision medicine will help identify which treatment will work for an individual.

And the paper is not just by the PhD student, it also bears the name of the leader of an important team working on ME/CFS research, a team that is a major part of the OMF fundraising machine. It's from a team that most of us recognise as one of the best teams working on ME/CFS out there. This paper and others published by them provide an indication of the attitudes and effort of all of those people. I believe it is reasonable and even important to express concern when their papers indicate that they hold some wrong ideas.

 

[Hutan]

The promise of “big” biological data

This section has some nice content.

Omics technologies are now routinely employed in ME/CFS studies, offering numerous opportunities to explore potential biomarkers. Due to the hypothesis-generating nature of these omics datasets, study outcomes can vary based on several factors including the chosen omics platform, batch effects, sample collection methods, analytical instrument [34], storage and handling. Increasing the sample size is often considered one of the most effective strategies for minimising the influence of technical outliers and strengthening statistical power without indirectly introducing bias. However, this approach can be limited by practical constraints such as cost and data availability. In such cases, normalisation and batch effect correction techniques can be applied post-data acquisition to reduce variability [35], however, different techniques may change study outcomes. When sufficient quality control data or internal standards are available, technical variation can be measured and subsequently removed [36].

I would have liked to see some discussion here about issues specific to ME/CFS that might affect the quality of analysis, as these things are questions we often have about studies and would be useful for new researchers coming into the field. So, perhaps reprising the issue of mis-diagnoses (with biases perhaps varying by location). So, you might get a systematic difference between ME/CFS cohorts from two different clinics and it can be very hard to know if it is the result of differences in the type of patient included in the cohort or due to differences in storage of the samples or of they way they were analysed.

Also, the difficulties around collecting samples from people with severe ME/CFS, and how this can affect the quality of the samples (e.g. does it take longer to get the samples processed?).

Perhaps these things are discussed later.

Reference #34 is
Huang K, Thomas N, Gooley PR, Armstrong CW. Systematic review of NMR-Based Metabolomics practices in Human Disease Research. Metabolites. 2022;12(10).
It sounds interesting; it's by most of the authors of this paper. We don't have a thread for it yet.

 

[Hutan]

Machine learning concepts

This section seems to me to be a useful introduction to the topic

When a model is trained on more features than samples (a phenomenon known as the curse of dimensionality), it may become overfitted, meaning that the patterns learnt are too specific to the dataset. Consequently, the model may not make reliable predictions on new input, especially from different data sources. There are various methods to prevent overfitting

Classification tasks in ME/CFS

The classification pipeline includes data partitioning, data preparation, feature selection, model selection, training, and evaluation with a blind test set (Fig. 3). In ME/CFS, classification applications have focused on biomarker discovery and diagnosis. The main difference between these endpoints is that the biomarker discovery studies typically do not choose a single optimal model; instead, important features from all candidate models are considered as potential biomarkers.

Table 1 is a list of recent ME/CFS studies classifying big data. It is noticeable how small most of the samples are. The importance of validating the findings is stressed, by holding back a portion of the data for the validation test, and by replicating the finding using other data.