Health Rising’s Recent Recovery/Recovering Series

Health Rising’s recent Recovery/Recovering Series continues with – you guessed it – something entirely different. We’ve had vaccine recoveries, drug (rapamycin) recoveries, neuroplasticity + recoveries, cerebral spinal fluid patch recoveries, mast cell + recoveries, and now we have our first – but hopefully not last – artificial intelligence-aided recovery, which, interestingly enough, took a different thrust at ME/CFS than we’ve seen before. Check out Efthymios’s story below.

My ME/CFS Remission Story 

A data scientist, Efthymios used artificial intelligence to chart his pathway to recovery.

Prior to 1999, my health was stable, although I would get more easily fatigued compared to people of my age. Whenever I was on holiday, for example, I was the only one who would spend significantly more time resting – especially if I was staying in the sun for too long – or being constantly active.

My first overt ME/CFS symptoms began in 1999 (after stopping a medication called Finasteride due to side effects) and advanced slowly. By 2004, Ι had full-blown ME/CFS. 

My symptoms included fatigue after exertion, tinnitus, urticaria (rashes), early morning insomnia, poor memory, difficulty finding words, brain fog, dysphagia (difficulty swallowing), secondary hypogonadism (reduced sex hormone production), anorgasmia (inability to achieve orgasm), orthostatic intolerance (symptoms while standing), numbness in extremities, sudden sensorineural hearing loss (3 events), anxiety, depression, mood swings, light and sound sensitivity. 

Test results indicated I was also in a pre-diabetic state. I was found to be hypothyroid, with mildly elevated TSH levels, and having increased homocysteine (Homozygous MTHFR mutation). At times, I would find only mildly elevated liver enzymes.

I was slowly deteriorating month by month and over time, began to have suicidal thoughts. Despair, though, turned to a commitment to find what was going on.  

Advanced Data Analytics Results in Stabilization

My background is in Computer Science with an MSc in Knowledge Extraction. I’ve been working since 2000 in Data Mining, and later in machine learning and natural language processing consulting. Still able to do mental work, I turned to advanced data analytics to develop software in order to research my own personal case and ME/CFS in general.  

Realizing that my symptoms changed in severity from day to day, I decided to keep a log for a total of 434 days to try and identify common factors (symptoms, food taken, supplements, etc.) associated with good vs. bad days using machine learning. 

After my first results in Nov. 2012 identified which factors (in terms of supplements used, food taken, weather patterns, and life events) were helpful or not, I followed personalized DOs and DONTs. (Note that these may not apply to others). 

With regard to foods, for instance, the machine learning analysis suggested that a fat-free diet, whey protein, iron-rich, and foods containing gluten were problematic while zinc-rich and cholesterol-rich foods were helpful. Weather changes were also clearly affecting me and any abrupt weather changes (e.g. increase in humidity, drop in temperature) would increase my symptoms.  

Another finding was that substances that were negatively affecting the cytochrome P450 system – which is related to the liver –  such as bioperine and grapefruit would greatly induce my symptoms.  

The results suggested that the following supplements might be helpful:  Any supplement providing methyl groups such as methylcobalamin, MSM, 5-methyltetrahydrofolate, and Zinc. 

Incorporating these changes of DOs and DON’Ts, and limiting significantly my alcohol intake, resulted in putting a stop to my continued worsening of symptoms and set the basis for my improvement.

Artificial Intelligence Provides Breakthrough 

Machine learning

Efthymios’s use of machine learning and AI tools to find central factors in ME/CFS and related diseases allowed him to take his health to the next level.

In 2013, I began developing a software framework in order to better understand ME/CFS and related syndromes as a whole, with a first working version being implemented by the end of 2014. (See upcoming artificial intelligence blog for more). 

Around March 2015, machine learning methods identified the following research targets: endoplasmic reticulum stress, oxidative stress, bile acid and phospholipid metabolism disruption, methylation, glucuronidation, sulfation, and detoxification pathways among others. 

These results provided a much clearer number of interventions to aim at. They focused on ameliorated endoplasmic reticulum and oxidative stress, support for methylation and detoxification phases, and amelioration of the inflammatory response.

They included: 

  • TUDCA – I started with 250 mg and gradually upped the dose to 750 mg, providing 250 mg every 8 hours in order to ameliorate endoplasmic reticulum stress (and subsequent unfolded protein response events) and also support liver function. TUDCA, a supplement and bile acid derivative, is another “version” of a medication called Ursodeoxycholic acid aka ursodiol: the liver produces bile acids which are metabolized in the intestines into TUDCA. These bile acids digest fat-soluble vitamins, break down food and cholesterol, help fight off pathogens, and regulate our metabolism. In an energy-intensive process, most of the bile in our body ends up getting reused. (More on bile acids and the liver in the next blog on Efthymios’s work on artificial intelligence and ME/CFS.)
  • Methylcobalamin (B-12) (1000 mcg) – to support methylation)
  • Methylfolate (200 mcg) – to support methylation)
  • Taurine (500 mg) – to help in conjugation of bile acids via CYP7A1, regulate NMDA / glutamate excitotoxicity, and also upregulate PGC1-α (for mitochondrial biogenesis)
  • Vitamin C (250 mg) and N-acetyl cysteine (100 mg – I couldn’t tolerate more) – antioxidants to minimize oxidative stress  
  • Ubiquinol (100 mg) – provides intermediate metabolites for energy production
  • Zinc (25mg) – which has been low whenever I tested it
  • Selenium (50 mcg) – (combat oxidative stress)
  • Vitamin K2 (As MenaQ -100 mcg) – (combat oxidative stress)
  • Choline (500 mg) and important cofactors such as FMN (Flavin Mononucleotide) (10 mg) (combat oxidative stress)
  • D-Limonene (250 mg) twice a week to induce P450 (liver support and detox) (Caution: this may affect drug metabolism)
Bile ducts liver

Two of Efthymios’s top supplements (TUCDA and taurine) support liver and bile production. (The bile ducts leading from the liver are in yellow.)


My recovery took a lot of trial and error – mainly in order to find which antioxidants were working – and happened gradually. I found that for my specific recovery, TUDCA, choline, taurine, and managing oxidative stress were particularly important.   


  • Efthymios Kalafatis, a data analytics specialist, put his skills to good use when he came down with ME/CFS in the early 2000s.
  • His use of machine learning to understand his personal patterns regarding diet, activity, supplements, symptoms, etc. allowed him to stabilize his health.
  • The analysis suggested, for instance, that a fat-free diet, whey protein, and iron-rich foods and foods containing gluten were problematic while zinc-rich and cholesterol-rich foods were helpful. Methylcobalamin, MSM, 5-methyltetrahydrofolate, and zinc were helpful supplements. 
  • In 2014, another machine learning analysis that incorporated factors from a range of ME/CFS and ME/CFS-like diseases broadened his approach. Now he was looking for broad disease-impacting factors that could conceivably apply to everyone with ME/CFS.
  • The analysis identified the following research targets: endoplasmic reticulum stress, oxidative stress, bile acid and phospholipid metabolism disruption, methylation, glucuronidation, sulfation, and detoxification pathways. 
  • Efthymios’s new protocol which focused, in large part, on reducing liver stress included a supplement called TUDCA, as well as choline, taurine, and antioxidants to reduce oxidative stress. (See the blog for a complete list). 
  • He also avoided any food with soy, glutamate, palm oil, gluten, sugar, simple carbohydrates as well as whey protein, and did not have a high protein intake.
  • He started feeling somewhat better in the first month and by 6-7 months he was fully recovered. 
  • He even went the extra mile – purposefully crashing himself and then adding the protocol back in bit by bit to see which parts of it were most effective.
  • Next up – a blog digs into Efthymios’s use of artificial intelligence and its surprising ability to uncover factors that have only recently shown up in ME/CFS research.

I ultimately went from a functional capacity level of about 4 (on a 1-10) scale to a full recovery. If my memory serves me well, I started feeling somewhat better in a month. By 6-7 months I was able to finally accept that I had gotten out of the “vicious cycle” of ME/CFS. I am now in full remission – including being able to do any amount of exercise – as long as I follow a number of interventions.

They include the supplements I mentioned (which I now cycle: TUDCA – use for 3 consecutive months and stop for 1 month. I check levels of folate, B12, and trace minerals, and when I reach a high level, I also stop for up to 1 month).

Also, the avoidance of any food with soy, glutamate, palm oil, gluten, sugar, simple carbohydrates as well as whey protein or increased protein intake. Loading on protein would always lead to a crash as well. (I hypothesized that if endoplasmic reticulum stress is a possible mechanism, too much protein may be putting an increased demand for protein folding and protein processing.)  

Any attempt to introduce many of the things I mentioned that I found to have a negative effect on me will eventually lead to symptoms and a crash, but my tolerance has increased over the years which means that I can now eat something containing – for example – monosodium glutamate for a number of days without crashing. 

In order to find what has worked for me, I purposely crashed numerous times in order to make sure that I was able to get into #MECFS and then get out of it. Many times I did not have the will to experience once again the numerous symptoms. But I had to make sure. This was done in order to see specific aspects of my regimen (e.g. do I need to constantly supply with methyl groups or not? Is TUDCA necessary?) and how much I am able to push myself. When a crash happens, I can get over it very quickly, within a couple of days at most.

(In 2017, an additional type of analysis named “Network Analysis” also identified research “hotspots” and further confirmed several of the findings from 2015, because it included concepts related to phospholipids, bile acid metabolism, peroxisome dysfunction, and many liver-related concepts such as Non-Alcoholic Fatty Liver disease (NAFLD) and hepatotoxicity.

More on that in the next blog on my efforts to use artificial intelligence to understand and find treatments for ME/CFS and related diseases. 

AI to the Rescue? AI-Driven ME/CFS Project Unearths Clues Ahead of its Time

Find Efthymios Story Intriguing? Please Support Health Rising and Keep the Recovery Stories Coming 





Like the blog you're reading? Don't miss another one.

Get the most in-depth information available on the latest ME/CFS and FM treatment and research findings by registering for Health Rising's free  ME/CFS and Fibromyalgia blog here.

Stay on Top of the News!

Subscribe To Health Rising’s Free Information on Chronic Fatigue Syndrome (ME/CFS), Fibromyalgia (FM), Long COVID and Related Diseases.

Thank you for signing up!

Pin It on Pinterest

Share This