Machine Learning-assisted Research on ME/CFS

I just realised another possible connection with microglia activation (VanElzakker). LXR was identified by Network Analysis in May 2017 :

https://www.ncbi.nlm.nih.gov/pubmed/25889344

 

Latest results : Hypoperfusion which is researched by Dr VanElzakker gets a very high ranking along with rbainstem, vagus nerve and cholinergic signalling.


I am currently looking at the connections of hydroxyproline and proline metabolism in general :

 

I am currently working in identifying more subgroups of issues responsible for ME (hypothesis). @wigglethemouse i haven't forgotten to see how mast cell actication could be fitting the greater puzzle of ME but please read below.


There are many of these issues that have been identified by researchers lately, two examples are Glycogen Storage Disease and Iron metabolism as identified by Birch et. al. As discussed i am using a custom-made Information Extraction system that looks at millions of PUBMED articles that tries to identify most interesting -in the context of ME- research papers.

This is one that i found lately, which discusses about mevalonate (aka Isoprenoid) pathway, a topic which was identified previously from Network Analysis. You may find several posts regarding the mevalonate pathway on PR as well.

Interestingly, some well known viruses downregulate the mevalonate pathway thorough induction of interferon : https://mbio.asm.org/content/8/4/e00576-17

The paper selected by the IE System can be found here : Note that it was selected through topics that did not include "Chronic fatigue syndrome" :

https://www.ncbi.nlm.nih.gov/pubmed/26983655

Now there are some entries about right hemispheric dominance which sound strange. There are however many interesting parts in the text and i wish someone with knowledge could have a look (the ever existing problem) :

Glutamate, again (?)


note the mention on of Na,K-ATPase :

Here is where it gets more interesting. From another paper :

Note that the paper shown above involves specific patient groups (hypertension, diabetes mellitus etc).


This is an example on how the IE system can potentially connect some of the findings we are presented with.

 

This was a very small sample and very preliminary data. Most Glycogen Storage Disease mutations require an autosomal recessive pattern, i.e. both copies of gene affected. Birch et al will need to follow up with transcriptomics, proteomics and Glycogen testing to prove that the mutation is affecting the disease to be sure.

Now my turn to take a hard right turn . Given the recent CCI news, I remembered about the Perrin point and poor lymph flow again. According to @Hip roadmap document
https://mecfsroadmap.altervista.org/

More info on the Perrin point from @Hip here
https://forums.phoenixrising.me/thr...your-left-breast-diagnostic-for-me-cfs.56148/

Perhaps worth looking out for lymph drainage in your brainstem search...................

 

Just a digression, but has anyone here been helped by the Perrin technique? What is the general opinion of this approach? Thanks.

 

Very interesting @mariovitali . I read this on Wikipedia (for other readers)
https://en.wikipedia.org/wiki/Mevalonate_pathway

Interesting given the lipid and heme findings in metabolic studies, and Jonas Berquists interests in the steroid hormone abnormalities.

"The role of red blood cell deformability and Na,K-ATPase" is interesting. Amit Saha (of recent ME RBC studies) is on twitter (@babyy_po) if you wanted to share anything with him there.........

I also thought Na,K-ATPase is interesting in relation to the nanoneedle salt test i.e. the pumping out of sodium. It's talked about that it's the lack of ATP that could be the issue with patient cells on the nanoneedle, but that is only a theory and has not been proven yet. Could just as equally be an issue with this Na/K pump. Here is a link for people to read up on Na,K-ATPase. It's really quite interesting.
https://en.wikipedia.org/wiki/Na+/K+-ATPase

 

@wigglethemouse if Amit Saha is going to be more receptive if you tell him, please do so. I added Na/K pump for scoring to Machine Learning to see what kind of ranking it will get and to see what other connections it has. I am using several different techniques to see its connections but for now the most important connection of Na/K pump is glutamate. Here is the number of connections with other topics in descending order.

glutamate
lactate
glutathione
pyruvate
glutamine
acetate
nitric_oxide
liver
creatine_kinase
dopamine
ammonia
nmda
reactive_oxygen_species
acetylcholine
excitotoxicity
quinolinic_acid
digoxin
norepinephrine
magnesium
isoprenoid
arginine
adenosine_triphosphate

I see also a number of reasons on why magnesium may have a very important role (not a suggestion to supplement).

 

Magnesium is the one supplement that is instantly noticeable if we miss it. It is key to so many reactions.
Lower core body temperature probably doesn't help either

 

Done.

https://twitter.com/user/status/1155206616351821824

 

Dear all, i would like to provide a short update. I begin with a number of entries of potentially higly significant abstracts to ME/CFS from PUBMED (using an Information Extraction system) . I provide some of them below (cc : @wigglethemouse )

There are several things which do not seem right in the paper cited above. It may have interesting information however such as Na,K-ATPase and its relationship with Cell deformability. Here is an interesting dialogue of @wigglethemouse with a stanford researcher on Twitter, let's hope something comes out of this :

https://twitter.com/user/status/1154801435659214848

The paper can be found here (Sci-hub) : https://sci-hub.tw/10.1016/j.mehy.2006.01.036 . It has some really interesting figures which i cannot copy-paste here.


Interestingly SP1, EGF and HMGB1 (mentioned in the paper) have been also previously identified by Machine Learning as research targets. :

https://twitter.com/user/status/1139425430434459648


https://twitter.com/user/status/1029647934156091392


https://twitter.com/user/status/1060073883888484353



And we also have the following :

There are many more available, i am trying to implement a better scoring system which will help in better selection.


I am also very happy to say that @Ben H will be also helping out with the execution of additional machine learning techniques that may help for creating hypotheses to the origin(s) of ME/CFS. Thank you @Ben H !

 

Thus, vagus nerve and cholinergic agonists inhibit systemic inflammation by activating the noradrenergic splenic nerve via the alpha7nAChR nicotinic receptors. alpha7nAChR represents a unique molecular link between the parasympathetic and sympathetic system to control inflammation.

The immunemodulator I use is a alpha7nAChR agonist.

 

The alpha7 nicotinic acetylcholine receptor as a pharmacological target for inflammation
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2042938/

 

More presice, a positive allosteric modulator.

In other words , levamisole is a key for the a7nAChR backdoor.

 

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3162128/
Positive allosteric modulators as an approach to nicotinic acetylcholine receptor- targeted therapeutics: advantages and limitations

 

Here is an update with the latest results.


The latest runs suggest a central role for glutamate metabolism. Glutamate has been found to be elevated in ME patients from some ME/CFS researchers. Also, we have new entries for Nitric Oxide metabolism and signalling, L-Arginine and vasoconstriction.

Regarding L-Arginine, I was very happy to read that there is upcoming research on this area by Dr Mercedes Rincon,I will definitely be watching this one :

https://cureme.lshtm.ac.uk/researchers/our-collaborative-researchers/

Regarding Glutamate metabolism, an email was sent to Professor Bergquist for possible existence of hypoxanthine,xanthine and uric acid in cerebrospinal fluid. More specifically, this paper was sent to him :

In his very kind and immediate reply, he explained that he will look more into this once he is back from his vacations.

Here are the latest results, depicting the features selected by certain feature selection methods. You will find many known entries :

pns = parasympathetic nervous system
crh = corticotropin-releasing hormone






Hypoperfusion, brainstem and the vagus nerve are still being selected in the highest positions (VanElzakker). Observe how nitric oxide and vasoconstriction are part of some results. Entries relevant to inflammation appear frequently.
@wigglethemouse you may also see BDNF in the results . I still do not understand why norepinephrine and epinephrine are part of the results.

Regarding indoleamine and kynurenine : they continue to be identified by network analysis as important "pieces of the puzzle" however i postulate that the reason is ultimately glutamate metabolism and the connection with quinolinic acid (Dr Phair is aware of this connection) :

 

Good stuff @mariovitali . Can you explain the three groupings in your screenshot - not the results, but what are the groupings showing - Chi Square, Consistency Evaluator, Information Gain.

 

These are all feature selection methods. They are aimed to select features with high importance for the problem at hand. In this application we wish to identify which features (where a feature can be any biological concept) are relevant to ME symptoms.

I tend to use several tools and combine different methods. The one shown is from a Data Mining tool called "Weka".To anyone interested, here is a description of some feature selection methods :

https://www.matec-conferences.org/articles/matecconf/pdf/2016/05/matecconf_iccma2016_06002.pdf

 

@mariovitali In Alain Moreau's Stanford Symposium presentation he showed this slide where 3 out of 4 twin families showed a mutation in SLC1A2 related to Glutamate clearance in the brain. Glutamate keeps popping up!

 

@wigglethemouse Thank you for this "catch". I think it is also interesting that Na,K ATP-Ase is associated with SLC1A2 aka EAAT2 aka GLT-1 :

https://en.wikipedia.org/wiki/Excitatory_amino_acid_transporter_2

https://academic.oup.com/cercor/article-abstract/29/8/3331/5107846?redirectedFrom=fulltext

EDIT : The above possibly refers to mice

 

Dear all,


Here is the latest update. cc @wigglethemouse , @Perrier @Ben H , @rvallee , @ScottTriGuy


I believe that significant progress has been made ( meaning for symptoms related to myself ,so n=1 !). In order to better understand how is all connected i've been crashing myself purposely to perform root cause analysis.

For this reason - with a little help from technology - i am trying to identify how different promising research targets (hypoperfusion, sepsis, inflammatory response,hypoxia, glutamate and others etc) fit the picture. I would like to provide a small example of the work i've been doing. Please see below :



The map shows a learned network of interactions between these topics using machine learning(the best way to put it). Observe how inflammatory response is connected with sepsis and the gut / albumin. Why the connection with albumin ? Perhaps because inflammation lowers albumin levels. My first question is : Do we find hypoalbuminemia or even hyperalbuminemia to ME patients ?

We next see the connection with "ckd" which is chronic kidney disease, a topic which i have mentioned many times. Here is where eGFR may be coming to the picture. Do ME patients are found with significantly lower eGFR than Healthy controls? We move through ckd to angiotensin, endothelin to vasoconstriction which leads (?) to hypoxia and hypoperfusion (?).

Please note, this is just an example of the tools that we already have and no one is currently using. These tools are supposed to work with the help and guidance of medical professionals.

I hypothesise the following for further research :

a) The Inflammatory response is a key piece of the puzzle. Proposed topics here are IL-6 and TNF-α
b) The same applies for glutamate/kynurenine metabolism. The common factor of these two is excitotoxicity and tinnitus may be related to vasocontriction / hypoxia/ excitotoxicity or even hypoperfusion
c) Metabolism of Nitric oxide : more work is needed here

I can induce the perfect storm if i eat too much protein (whey protein has a negative effect too), Parmesan cheese (contains glutamate) and many other glutamate sources. Please have a look here :


https://www.msgtruth.org/what-foods-should-i-avoid


and also the "Umami" database :


https://www.umamiinfo.com/umamidb/search/

I cannot really suggest you try a glutamate-full diet because it may induce a severe crash (if the theory is applicable). Whatever i write here is not a suggestion to try but for informational purposes only.

I move on to -yet- another available tool that is not currently being used to speed things up. I will provide an example based on the above. As discussed, inflammatory response may be a key topic related to ME pathology. Imagine being a researcher and you wish to see which compounds have an effect on ameliorating the inflammatory response. At the moment of writing, "inflammatory response" has 72479 entries on PUBMED. Using an Information Extraction tool, we are getting the following excerpts which contain (almost) exactly what we need.

The list above is not the complete output. The extraction of this information from 72479 PUBMED abstracts required a total runtime of 1 minute and 24 seconds.