Application for RAMSAY Awards 2019 - Machine Learning-assisted Research on ME/CFS

[mariovitali]

Dear All,

I wanted to inform you about my application to SolveCFS for a RAMSAY award funding and the final decision. The title of the application was :

"The use of artificial intelligence and network analysis methods to identify mechanism(s) responsible for Myalgic Encephalomyelitis/Chronic Fatigue Syndrome"


I present the introductory text :

In this project, artificial intelligence and network analysis methods will be used to identify possible mechanism(s) responsible for Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS), a multi-system disease with unknown cause(s). To date, research efforts have focused on singular, or narrow, areas of research, creating silos of research that lack the ability to investigate the interplay of symptoms, genes, biological pathways, and co-occurring diseases.

Most of our scientific knowledge regarding these relationships is published as text, which is difficult to analyze through traditional statistical methods. To better understand the greater mechanism of this complex disease, a bespoke software framework will be used that employs advanced analytical techniques and consists of a number of software programs that work together to collect millions of published abstracts from an online database, rank them and identify most promising research topics according to their relevance to ME/CFS.

The main goal is to identify possible mechanism(s) for disease and potential subgroups of people with ME/CFS. This will be achieved by:

1. 1) identifying topics related to symptoms, genes, biological pathways, and co-occurring diseases from published abstracts directly or indirectly related to ME/CFS;
   
   
2. 2) identifying previously unexplored associations between symptoms, genes, biological pathways, and co- occurring diseases of ME/CFS; and
   
   
3. 3) generating hypotheses of possible mechanism(s) for and potential subgroups of ME/CFS.
   

In preliminary work to lay the foundation for this software framework, millions of abstracts were collected (not only directly related to ME/CFS) published since 01 Jan 1970. With each update (incorporating the latest publications), the framework identifies research topics and ranks them according to their relevance to ME/CFS symptoms and pathology; initial use has successfully identified a number of hypotheses that were later confirmed by researchers using traditional methods.

This project will begin by collecting the most recent publications, identifying previously unexplored associations between research topics and then assessing these associations for biological rationale. These steps will be repeated on a monthly basis for eight months to incorporate all new publications into the framework. Through these findings, informed hypotheses on possible mechanism(s) for ME/CFS and on potential subgroups of patients with ME/CFS will be generated.

The application was scored by two reviewers for Significance, Innovation, Approach and Overall Impact. I believe that it is not right to share the comments of the two reviewers. They were invaluable in making me understand the pros and cons of the solution that was built over the past 6 years and the difficulties i had in communicating the capabilities of this framework.

Waiting for the "verdict" was a painstaking process (the application was submitted in July). After all, i waited for this moment since 2013.

I was informed yesterday that the application was not successful. I do not -by any means- wish to criticise this decision. I did not have the opportunity to "defend" my application with the reviewers but i do understand that this process may not be applicable. The news took a toll on me. I was prepared for a negative outcome but i also had positive thoughts. Creating such framework was not easy. I had to work very hard and this effort had several negative repercussions on my personal life and also financially-wise as i had to work for many - many hours.

I thought that the right thing to do is to finally give up and move on with my life. I then reminded myself that i am symptom-free because i didn't give up and instead persisted in finding a solution for me.

As i mentioned in older posts, the great thing about forums is that they leave a written record of whatever is being said. Here are my takeaways :

- I would like to particularly Thank @ScottTriGuy for his help and tremendous effort to make this work heard. Also Jane, @wigglethemouse and @Ben H .

-I stand strong in believing that AI will be a game-changer for putting the pieces of the puzzle together

-I will do anything possible to have results of this work being used by other researchers helping them to better understand the bigger picture. I have ready specific blood tests, compounds and also identified potential sub-problems (e.g Cysteine catabolism, Glutamate metabolism) that may be responsible for the perfect storm we are seeing. Unfortunately under these circumstances this task will not be easy.

I will never quit. Ever. Until we find a solution.



Stay strong,

"Mario"

 

[Trish]

Thank you @mariovitali for the dedicated effort you have put into this project for 6 years. I am glad it has helped you find a solution for your own health, and hope your improved health is sustained.

I sympathise with your disappointment over the outcome of your application for funding. It's a tough old world out there, and so many good research ideas don't get funded.

I'm pleased you are persisting in passing on what you have learned to other researchers.

I understand you don't want to share what particular approach worked for your health, as it might encourage others to try it for whom it is not suitable and could potentially cause harm.

I wonder if you have communicated your own treatment and outcome to any researchers who might be able to investigate it further to see if it could be applicable to other patients and could be tested in a clinical trial.

All the best to you.

 

[Andy]

Sorry to hear you weren't successful this time Mario, glad to hear though that you intend to persist in your efforts.

 

[ME/CFS Skeptic]

Thanks for your efforts Mario,

 

[mariovitali]

Thank you @mariovitali for the dedicated effort you have put into this project for 6 years. I am glad it has helped you find a solution for your own health, and hope your improved health is sustained.

I sympathise with your disappointment over the outcome of your application for funding. It's a tough old world out there, and so many good research ideas don't get funded.

I'm pleased you are persisting in passing on what you have learned to other researchers.

I understand you don't want to share what particular approach worked for your health, as it might encourage others to try it for whom it is not suitable and could potentially cause harm.

I wonder if you have communicated your own treatment and outcome to any researchers who might be able to investigate it further to see if it could be applicable to other patients and could be tested in a clinical trial.

All the best to you.

@Trish , Thank you so much.

I was hoping to do that, being under different circumstances (in case that the award was given) as this would have made researchers more receptive. SolveCFS may help me in contacting research labs and if such thing takes place, it would be a great start.

There are many things to share . SolveCFS will be providing a mobile application for ME patients to track their symptoms. Through the CureME team, i managed to inform SolveCFS that i did the same kind of tracking in 2011 and shared with them some Best practices plus some thoughts on having people continually use the application. One of the most difficult parts in keeping a log ( i had to do this for 434 days) was not giving up but continuing each and every day to log symptoms.

Certain insights were extracted such as being on a fat-free diet induced my symptoms (wasn't aware of that) and how gluten would induce brain fog (which i kind of knew). The other amazing thing i noticed was how weather variations induced my symptoms with greater being humidity. I had to use specialised hardware to receive data as these changes were happening :

https://forums.phoenixrising.me/threads/me-better-when-abroad.53988/#post-897595

My hypothesis is that high humidity makes worse the inflammatory response that takes place. I strongly believe that inflammation is a key aspect for many ME patients and their symptoms.

 

[ringding]

I'd just like to echo the comments above, particularly Trish's. Thank you for your persistence and I'm very pleased that you are healthy and intend to continue your efforts.
The whole experience is understandably personal for you and yet you seem to have maintained your objectivity. Not easy by any means. I tip my hat to you.

 

[Adrian]

-I stand strong in believing that AI will be a game-changer for putting the pieces of the puzzle together

One of the issues with AI/ML being used is the quality of the data being fed into the algorithms. For example, looking for connections over multiple abstracts may pull out something useful but it could easily be perturbed by bad data (i.e. too many small studies which may not have been done rigorously). ML (using an L2 norm (i.e. least squared cost function) is particularly sensitive to mislabeled data - although I don't know what algorithms you are using.

Hence in other fields when looking at applying ML to problems it is important to address the data quality issues.

 

[mariovitali]

One of the issues with AI/ML being used is the quality of the data being fed into the algorithms. For example, looking for connections over multiple abstracts may pull out something useful but it could easily be perturbed by bad data (i.e. too many small studies which may not have been done rigorously). ML (using an L2 norm (i.e. least squared cost function) is particularly sensitive to mislabeled data - although I don't know what algorithms you are using.

Hence in other fields when looking at applying ML to problems it is important to address the data quality issues.

Agreed, there are inherent problems. Keeping in mind that the data originates from PUBMED abstracts we also have issues such as publication bias. I decided to go ahead and build the system despite these issues and see what would come out and whether findings make sense.

Since you appear to have knowledge about these technologies, here is a short description :

a) For text analysis, Information Extraction and Natural Language Processing are used
b) matched medical entities are transformed to a numerical representation, suitable for further analysis
c) Machine Learning algorithms then try to identify which medical entities are associated with the cluster of symptoms we are finding to ME/CFS. The libraries/software used is SCIKIT-Learn, WEKA, R,KNIME, H2O and others.
d) Network Analysis attempts to find entities, central to ME/CFS-related topics

 

[wigglethemouse]

The other amazing thing i noticed was how weather variations induced my symptoms with greater being humidity. I had to use specialised hardware to receive data as these changes were happening :

@mariovitali Have you come across any API's or tools to extract historical weather data from the internet? I'd love to track data because like you I've noticed a correlation but have no data to back it up.

 

[mariovitali]

@mariovitali Have you come across any API's or tools to extract historical weather data from the internet? I'd love to track data because like you I've noticed a correlation but have no data to back it up.

I never used an API but have been using wunderground. Here is an example :

https://www.wunderground.com/history/weekly/gr/spata/LGAV/date/2019-10-8

To be sure about the effects of weather on my self, i had to buy hardware. So i ended up installing a Kestrel 5000 series weather station at my balcony which measures wind direction, humidity, temperature and many others. it was important to track weather changes as they happened.

https://kestrelmeters.com/

This was a very cool setup. I had the weather station connected to my cell phone through bluetooth and could see the immediate consequences that wind gusts had on the pain i was experiencing or other symptoms.

cc : @Perrier

Here are some notes from how my body responded.Please keep in mind that these notes are applicable to me but you may find that the patterns also apply to you :

As discussed, i understood that tinnitus was a key aspect of my symptoms. I later understood that tinnitus signified a "crash" (and was part of my vicious cycle)

1) I realised that tinnitus would happen before major weather changes : Humidity, wind speed, temperature drop
2) Symptoms would be more intense with combined changes. Example : High humidity weather and high south winds. North winds were better vs South winds
3) Especially when the weather was humid, i could feel the pain on active sites of inflammation (e.g my periodontitis and certain points where the problem is bigger). Also my liver area and spleen.
4) Ideal weather conditions associated with far less symptoms : No wind, no rain, humidity at around 50%. Places with high weather variability would be the hardest ones for me (e.g the UK )

This was very important information because i was able to have a feedback mechanism to optimise my regimen. So using the weather station i would see -given weather conditions- how many symptoms i was experiencing. If my symptoms would be less on a high humidity day then that meant that my current regimen was doing something beneficial.

Then we have the connection with food. I knew that some foods would also immediately induce tinnitus :

-Whey protein
-Parmesan cheese
-Protein Bars

The common factor of these is high glutamate content. Another thing i am looking at is foods with high Polyamine content :

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3022763/

I am not sure about this yet, but i believe that high content of polyamines may not be beneficial. It's been identified algorithmically but -as discussed- the algorithms do not tell us if what is identified is good or bad for our symptoms. My hypothesis is that Polyamines are important because their connection with Diamine Oxidase (DAO) functioning and GABA production but i need to work more on this.

Glutamate is excitotoxic. This is why i would be very interested to know @Andy what Dr Van Elzakker has to say about this. Can we find excitotoxic lesions anywhere? Can we measure it in vivo?

 

[Andy]

Glutamate is excitotoxic. This is why i would be very interested to know @Andy what Dr Van Elzakker has to say about this. Can we find excitotoxic lesions anywhere? Can we measure it in vivo?

We did discuss glutamate and glutamine but I'm afraid we didn't get to those specific questions, not enough time for all the questions suggested.

He did suggest that we might do another video at some point, so I may well get the chance to address the questions that I didn't ask then, or if you didn't want to wait for that then perhaps contact him direct?

 

[mariovitali]

We did discuss glutamate and glutamine but I'm afraid we didn't get to those specific questions, not enough time for all the questions suggested.

He did suggest that we might do another video at some point, so I may well get the chance to address the questions that I didn't ask then, or if you didn't want to wait for that then perhaps contact him direct?

This might be just my impression, but he was not very receptive to my tweets mentioning him. I think it would be better for someone else to get in contact with him just to be on the safe side.

 

[mariovitali]

@Andy

cc : @Perrier , @wigglethemouse

I believe this should be one more interesting question for Dr Van Elzakker. Vitamin K metabolism which has been identified through Network Analysis since 2017 :

Vitamin K-dependent gamma-glutamylcarboxylation: an ancient posttranslational modification.
Bandyopadhyay PK1.
Author information
1
Department of Biology, University of Utah, 257 South 1400 East, Salt Lake City, Utah 84112, USA.
Abstract
The vitamin K-dependent carboxylase carries out the posttranslational modification of specific glutamate residues in proteins to gamma-carboxy glutamic acid (Gla) in the presence of reduced vitamin K, molecular oxygen, and carbon dioxide. In the process, reduced vitamin K is converted to vitamin K epoxide, which is subsequently reduced to vitamin K, by vitamin K epoxide reductase (VKOR) for use in the carboxylation reaction. The modification has a wide range of physiological implications, including hemostasis, bone calcification, and signal transduction. The enzyme interacts with a high affinity gamma-carboxylation recognition sequence (gamma-CRS) of the substrate and carries out multiple modifications of the substrate before the product is released. This mechanism ensures complete carboxylation of the Gla domain of the coagulation factors, which is essential for their biological activity.

Is it possible that problematic functioning of genes related to Vitamin K metabolism affect glutamate modification / catabolism? Could a perfect storm be a combination of Vitamin K-dependent functioning along with other problematic genes responsible for Glutamate metabolism/catabolism such as GAD1, DAOA ?

Note that warfarin affects Vitamin K absorption (snapshot from SolveCFS website, slide from my presentation at EUROMENE) :







Network Analysis, generated in 2017 :







Perhaps i should have written "glutamate catabolism" below instead of "apoptotic cell clearance impairement". Note also that the 3 SNPs of this severely ill patient have to do with warfarin response.

 

[ScottTriGuy]

@mariovitali - It is so frustrating trying to get funding for ME research, and the Ramsay awards seemed like such a good fit - no doubt it is discouraging not receiving funding, especially when so many patients are suffering and so much research needs to be done - since there is so little funding available generally, I imagine there were a lot of applications competing.

It takes grounded wisdom to able to put everything - including your considerable time, energy, money - into the larger context and take away feedback from the reviewers as learning to take forward, that speaks to your openness, resilience and commitment to the cause.

Thanks for continuing to push your research forward, in spite of the seemingly Sisyphean up hill path, because what you've uncovered and your self-experiments and your recovery, have great value that needs to be delved into deeper. Sooner than later.

 

[Midnattsol]

Remember glutamate in itself is not bad, it's when the body cannot regulate it property it becomes problematic. And when that happens it can increase oxidative stress which in itself can lesd to dysregulation of this system. That's one viscious circle. How about glutathione regulation? It is needed to have proper balance between Glutamate and GABA. Also, remember glutamate from food does not necessarily cause problems in the body. The foods you mention also include potentially problematic compounds for some (like casein, lactose, sugar and whey are all known to increase inflammation in susceptible individuals).

That warfarin gives symptoma of dysautonomia could be due to blood pressure issues? It's a blood thinner after all. Vitamin K is needed for blood clotting, so persons on warfarin sometimes need to adjust their dose if their dietary intakte of vitamin K changes. Blood pressure falls for me are a PEM symptom, and taking a drug that would decrease blood pressure even further I guess would give me symptoms.

Gah, I'm a bit too tired to write in a proper manner what I would like to say, will perhaps try again later. Good work, tho!