Machine Learning-assisted Research on ME/CFS

@mariovitali

I don't know anyone specific to suggest, but would environmental medicine associations be of some help?

My understanding is, these are MDs, and possibly naturopathic doctors as well, who look at our environment and its effects on our health. Some of their focus is liver function - detox pathways. There are labs that these physicians work with to test their patient's liver function, toxicity from heavy metals, chemicals etc. Perhaps some of these types of groups, and/or labs might be of help.

Another thought, I'm not entirely sure, but can concepts, hypotheses etc. be patented? If so, that might be a way to retain ownership.
 
Last edited:
@wigglethemouse @wastwater


I had a look at full DNA data of patients that i have. There are two patients out of 7 that were found to have SNPs in Dicarboxylic Aminoaciduria of uncertain significance :


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

From Wikipedia :

Dicarboxylic aminoaciduria is a rare form of aminoaciduria (1:35 000 births[2]) which is an autosomal recessive disorder of urinary glutamate and aspartate due to genetic errors related to transport of these amino acids.[3] Mutations resulting in a lack of expression of the SLC1A1 gene, a member of the solute carrier family, are found to cause development of dicarboxylic aminoaciduria in humans. SLC1A1 encodes for EAAT3 which is found in the neurons, intestine, kidney, lung, and heart.[3][4] EAAT3 is part of a family of high affinity glutamate transporters which transport both glutamate and aspartate across the plasma membrane.
Click to expand...


The SNPs are :"


rs559846052 MAF =T=0.000799/4 (1000 Genomes)

rs2229885 MAF = C=0.007388/37 (1000 Genomes) but appears more frequent (around 1%-3%) in other GWAS
 
Dear All,

I would like to provide an update -after a very long time- regarding my effort in researching ME using advanced analytical methods.

1) Regarding my joint effort with the CureME team, we are in the final stage of reviewing a paper for submission that explores how ME patients use supplements and medications. This work was also presented at the American Public Health Association (APHA) because it used novel methods in order to work with unstructured data (ie text)

More details can be found here : https://apha.confex.com/apha/2019/m...lename=2019_Abstract438850.html&template=Word


2) At last, an ME researcher has shown interest on the "signals" that Machine Learning (ML) has been giving for the past 5 years. As i was asked to keep full confidentiality i am not able to give details of this joint effort other than the fact that there is systematic work for the past 2 months to assimilate what ML has identified as potential areas of research and attempt to put pieces of this puzzle together. The key difference here is that -for the first time- someone with medical knowledge is able to guide the process, pose questions to the ML system and assess the validity of its output.

3) Just a few days ago the system identified a compound that could -according to the hypothesis generated by ML- normalise the nanoneedle response in stressed cells. I will keep you updated as soon as i have more news.


Finally, i wanted to comment -and these are only my personal views- on a subject that is not well received by ME patients : The mental manifestations such as depression and anxiety. I was looking at the responses in the following thread :


https://www.s4me.info/threads/perfe...-a-systematic-review-2020-cherry-et-al.18106/


I believe that it is very important that we look at the mental manifestations as well. This is by no means a statement or a hypothesis that we got ME because we are mentally unstable. Far from it. But it could be possible that high anxiety, ADHD, depression can be attributed (ie caused) by biological mechanism(s) that are also responsible for getting ME. Many readers may now be asking "Does this mean that highly anxious personalities are more prone to ME?"

This is not the case. There is research that shows that depression could be attributed to a subset of patients to mitochondrial dysfunction. Some examples :


https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5997778/

https://www.quantamagazine.org/mitochondria-may-hold-keys-to-anxiety-and-mental-health-20200810/

https://www.annualreviews.org/doi/abs/10.1146/annurev-clinpsy-082719-104030?journalCode=clinpsy


We have to ask many questions going far back from when each individual got ME. If, for example an ME patient discloses that he/she has always been highly anxious or always has been very sensitive to criticism then this would definitely be something to look at further.
 
mariovitali said:
2) At last, an ME researcher has shown interest on the "signals" that Machine Learning (ML) has been giving for the past 5 years. As i was asked to keep full confidentiality i am not able to give details of this joint effort other than the fact that there is systematic work for the past 2 months to assimilate what ML has identified as potential areas of research and attempt to put pieces of this puzzle together. The key difference here is that -for the first time- someone with medical knowledge is able to guide the process, pose questions to the ML system and assess the validity of its output.
Click to expand...

I am curious as to what level of guidance/refining is possible? (I'm unfamiliar with the details of this process)

In terms of biological modelling, I'm not so interested in generating hypothetical ideas from scratch, but to refine hypotheses where the precise signalling relationships may be difficult to pin down due to complexity. Specifically, generating directional signalling pathways, based on certain fixed assumptions, but using such a system to discover/reveal other high probability pathways based on primary research. And in particular, identifying feedback loops that could perpetuate such signalling. An additional difficulty is also recognising that the same receptors/ligands/pathways can have specific variations depending on the type of tissue/microenvironmental conditions and thus I'm guessing would require some clever tricks to control the inputs to make sure they are relevant. Is this sort of approach possible?

mariovitali said:
We have to ask many questions going far back from when each individual got ME. If, for example an ME patient discloses that he/she has always been highly anxious or always has been very sensitive to criticism then this would definitely be something to look at further.
Click to expand...

The problem with this is recall bias and selection/participation/Hawthorne effect biases. Those sorts of studies are unreliable unless conducted in a long-term prospective manner with screening before participants come down with particular illnesses. Unfortunately, given the incidence, the initial enrolment sample sizes need to be very large for statistical validity. For example, with incidence of 0.015% person-years, an enrolment of 300,000 participants(!) would lead to 45 cases to study.
 
Snow Leopard said:
I am curious as to what level of guidance/refining is possible? (I'm unfamiliar with the details of this process)

In terms of biological modelling, I'm not so interested in generating hypothetical ideas from scratch, but to refine hypotheses where the precise signalling relationships may be difficult to pin down due to complexity. Specifically, generating directional signalling pathways, based on certain fixed assumptions, but using such a system to discover/reveal other high probability pathways based on primary research. And in particular, identifying feedback loops that could perpetuate such signalling. An additional difficulty is also recognising that the same receptors/ligands/pathways can have specific variations depending on the type of tissue/microenvironmental conditions and thus I'm guessing would require some clever tricks to control the inputs to make sure they are relevant. Is this sort of approach possible?
Click to expand...

Good question. The patented methodology does not employ directional signalling pathways but first identifies all relevant elements in each pathway but also biological events (e.g ER Stress). What you mention has been implemented -to an extent- by an ME patient (i do not know if he would like his/her name to be disclosed) and i may ask his/her help to do exactly what you suggest.

From the beginning @Snow Leopard i've been trying to put on the table what we are finding in ME (e.g increased lactate, C1Q antibodies, cardiolipin antibodies etc) and then have a system identify where all ends meet. This is what the framework that was developed is doing. Unfortunately, i feel that there is no systematic way from any ME research team trying to achieve the same result.

Snow Leopard said:
The problem with this is recall bias and selection/participation/Hawthorne effect biases. Those sorts of studies are unreliable unless conducted in a long-term prospective manner with screening before participants come down with particular illnesses. Unfortunately, given the incidence, the initial enrolment sample sizes need to be very large for statistical validity. For example, with incidence of 0.015% person-years, an enrolment of 300,000 participants(!) would lead to 45 cases to study.
Click to expand...

Understood. But we have to try given the limited resources. We may generate noise but also find some signals among this noise. It is important to know however that given the circumstances we expect to see noise and proceed with a lot of caution.

(Edited for clarity)
 
Last edited:
Not sure if I posted this here. For those who remember it, Network Analysis identified in May 2017 the importance of peroxisomes and choline deficiency (Snapshot from Phoenix rising):

@Hutan @Snow Leopard @Trish @Andy

Screen Shot 2023-01-14 at 13.31.17.png



5 years later, a metabolomics study identified the following :



peroxisomes.png


There are many more examples like the one above. I tried many times to ask from organisations and researchers to use this analytical framework,at no cost. No one decided to give it a try in order to connect the necessary dots . Bhupesh Prusty recently tweeted about issues with the Liver. So did Dr Karl Morten.

Unfortunately researchers have a bias toward their subject of research. I understand that, but this will not help us find a solution.

If you still wonder how it is possible that a disease of this kind is a total mystery to this day : Add to the possible reasons the "selfish gene"

As always -although some times I acted quite defensive- I am open to any thoughts and criticisms. What is the best way to move this work forward? I only imagine, if I had the necessary help from medical experts - I know nothing about medical science- where would we be.
 
Hi @Trish

I am doing fine, Thank you. I prefer not to get into details as to what it happened with researchers. There were some good moments but many more bitter ones.

This will change very soon. I think. I feel very disheartened that it took so many years for someone even to listen, let alone giving credit for one's work. Wishing you all the best as well.

Maybe it's the selfish gene talking now. Anyway, if someone else has any comments I would be more than happy to hear them.
 
mariovitali said:
Hi @Trish

I am doing fine, Thank you. I prefer not to get into details as to what it happened with researchers. There were some good moments but many more bitter ones.

This will change very soon. I think. I feel very disheartened that it took so many years for someone even to listen, let alone giving credit for one's work. Wishing you all the best as well.

Maybe it's the selfish gene talking now. Anyway, if someone else has any comments I would be more than happy to hear them.
Click to expand...
In Canada (saying from Mohawks): too many Chiefs here not enough Indians.

Selfishness at the expense of lives, and suffering, and a very vulnerable population. And many young people who have died and continue to die.It is appalling that folks are not willing to see what others have to offer.
 
@mariovitali I've only ever tried the basics of ML on blood data, with training and validation, so I don't have much experience. That data showed clear separation so it was kind of overkill for ML as it turned out. I have a couple of questions, apologies if they've been asked before.

Do you/can you re-run your system annually and if so how (if at all) have the highlighted genes/pathways changed given more recent literature?
When you have a hit such as TYRO3 above, can the system give an indication of high-weighted articles so one can manually review its suggestions, or is it "simply" building up a huge network of word/term/phrase linkages across the dataset that remain a black box, and you only have eg "go look at TYRO3"?

(As much of a fan of computer automation as I am, I'm still reading the research manually and following paths of interest as they come up, ie first and second order references and cited-bys).
 
@SNT Gatchaman

Do you/can you re-run your system annually and if so how (if at all) have the highlighted genes/pathways changed given more recent literature?
Click to expand...

Yes, this is possible to be done and more often if needed

When you have a hit such as TYRO3 above, can the system give an indication of high-weighted articles so one can manually review its suggestions, or is it "simply" building up a huge network of word/term/phrase linkages across the dataset that remain a black box, and you only have eg "go look at TYRO3"?
Click to expand...

Yes but not only that. The system can help us look at the bigger picture. An example is shown below related to results identified by machine learning (in green), latest findings (red) and blue(targets identified by Maureen Hansons group in August 2023).

LXR+ABCA1+MERTK.png



In other words, one important aspect is the role of Vitamin K related genes (MERTK, GAS6, PROS1 and TYRO3 which is related to MERTK) and the potentially important role of apoptotic cell clearance to the pathology of ME/CFS and possibly LongCOVID.
 
You must log in or register to reply here.
Back
Top Bottom