The article below from The Economist, Feb 24, 2023 hints that ChatGPT and similar machine learning programs being used in Medicine will likely kill people “By one estimate, this approach may have caused 10,000 deaths a year in Britain alone.”
Believe it or not, none is expected. Machine learning is not used. We use a very old school fuzzy logic expert system model. Additionally, we make microbiome samples available at http://citizenscience.microbiomeprescription.com/ and give the evidence trail as shown in this video.
This is actually a referral from a person from a previous blog, Another ME/CFS person has gone to Firmicutes!. He shared his experience (see that post) “My energy levels are perhaps the best they’ve been in the last 5 years. I’ve still got a very long way to go but the results thus far are promising!” and a friend decided to try
Back Story
I became symptomatic around 19, progressively got worse so that by the age of 23 I had fully crashed with classic CFS symptoms. Severe symptoms persisted for 5 years during which time I was unable to work due to unrelenting fatigue. Slowly got about 50% better through an extremely low stress lifestyle and dietary/food changes. I’ve tested positive for a few Lyme co infections, chronically low cortisol and pretty much anything else the chronic illness community tests for, I’ve tested and treated. I’m now 38 and my recovery seems to hang at about 40% of optimal capacity. I wake up unrefreshed and have lagging energy all day long. I have to live an extremely low stress life, if I don’t, my sympathetic system kick into high gear. I seem to have issues with histamines (though I can ingest them, I just get flushed, puffy and hot at times), and my fatigue gets profoundly worse around my cycle. I don’t have any significant digestive issues that I’m aware of.
Analysis
The first thing that I should mention is that I recall a study finding that the duration of ME/CFS does not impact the probability of remission. So 19 years with ME/CFS is not a factor.
Dr. Jason Hawrelak Recommendations came in at 99.9%ile, so no pro-forma general issues. That is not unusual, most labs on ME/CFS patients report normal. Looking at the Potential Medical Conditions Detected list, nothing of concern.
Looking at specific bacteria, a few bacteria stand out:
Mogibacterium at 100%ile — i.e. the highest level in over 1000 samples!
Lactobacillus is at 94%ile, a level many people would love to see (for ideological reasons). With ME/CFS, this causes the “D-Lactic Acidosis Claxton” to sound in my mind… D-lactic acidosis is deemed to be a common contributor to ME/CFS. A few past posts on this:
Looking at bacteria distributions we see a good pattern except with the rare bacteria which appear under represented. An ideal microbiome would have the same count in this range. This suggests a well established microbiome, perhaps with a touch of “inbreeding”.
Diving into the species of Lactobacillus interests me. We find just one species dominates, Lactobacillus rogosae. A 2018 review,”Occurrence and Dynamism of Lactic Acid Bacteria in Distinct Ecological Niches: A Multifaceted Functional Health Perspective“, states “All in all, no consistent marker for any pathology or a healthy state is simply defined by a specific proportion of Lactobacillus“. This strain being classified as a Lactobacillus has been challenged [1974] with the suggestion that they may belong to Propionibacterium, a family associated with Acne. We are back to the fuzziness of 16s lab software as well as challenge of RNA/DNA being exchanged between different bacteria.
We have irony here, because the friend was high in Firmicutes and we have 77% of the microbiome in this sample also being Firmicutes with heavy domination of several ones as shown in the Krona chart below
I am inclined to do the customary ones PLUS one just trying to reduce Lactobacillus to build the consensus.
The top items had one little surprise – Cadium! There is a source for this that is also known to be good for ME/CFS – dark chocolate! See Dark chocolate is high in cadmium and lead. Prior studies on ME/CFS found that dark chocolate/ cacao improves ME/CFS symptoms.
The avoid list containing many popular items claimed to help the microbiome (which it does in some cases). Some are obvious with high lactobacillus — i.e. avoid lactobacillus probiotics. lactulose is a key food for lactobacillus.
In short, avoid Lactobacillus and Bifidobacterium probiotics. We want to greatly reduce the Lactobacillus from the 94%ile – it is very likely that d-lactic acid from lactobacillus is responsible for many symptoms. Bifidobacterium INCREASES Lactobacillus which is the opposite of what we want to do. We want to get lactobacillus down as a first priority, later we look at increasing bifidobacterium once it is down far enough.
Q: Is there any reason you chose to focus on the lowering Lactobacillus, as opposed to the Bifidobacterium being low? Is it because the “bacteria deemed unhealthy” table is a more important focus to you than the jason hawrelak recommendations?
A: The medical condition of ME/CFS and Lactobacillus has a long connection to each other. Both “bacteria deemed unhealthy” and Hawrelak are general criteria. Being at the top of Hawrelak’s rating (99+% of people are worst), would imply no issues — you have issues.
Q: Why not focus on the Mogibacterium that’s in the 100 percentile?
A: We could — it was a factor included in the bacteria picked to modify. The list of items is here. Some people go after a single bacteria with a “all other factors be ignored”. For example, Fennel reduces Mogibacterium BUT it also increases Lactobacillus!! The AI algorithm attempts to balance the dozens or substances impact on dozens (or thousands) of bacteria. You are welcome to do it by hand for the 123 bacteria flagged and the several thousand modifiers.
Q: the last question is about interpreting Krona charts. The Krona chart doesn’t seem to provide standard ranges
A: This form of visual display will get extremely busy and confusing with ranges. If you attempt to draw a high range line it will sit over a different bacteria.
Q: So is there any easy way to know when a bacteria is high or low by looking at that chart?
A: Use the hierarchy chart. You can pick one set of ranges at the top of the page. Items that are high (by the selected ranges) are in blue, and low in pink. You can also hand pick bacteria to alter.
On My Profile TabChecking the checkboxes, then clicking Create… allows a hand picked set of targets.
So doing a microbiome test is like collecting the DNA from a bunch of people at a major airport and then asking: Which country did this person originated from according to their DNA. You will find some people that are good matches to an ethnic origin and some people that are “Heinz 57” aka “Mutts”. These people are unclassified — just like some living components are unclassified.
Let us look at my own DNA to illustrate the issue better. The same DNA file was used for ALL of the following charts. Why do they not agree? Simple — thing are done by matching patterns. The reference library determines where matches are done. Every provider use difference reference libraries. There is no universal reference for Human DNA, nor for the microbiome.
From 23 and Me
From Ancestry
and one more site, this one almost causes whip lapse!
FamilyTreeDNA
One site offers the choice of reference library to use and how to match
GEDMatch
GEDMATCH show where your DNA matches historical samples
I look very Irish here
When we drill down to the next level, we see different “species names”
From 23 and meAncestry
Wait — we are talking about where
The above patterns are based on matching to current populations. We really would like older populations. There is a site that does that! My True Ancestry. The same DNA file suggests more UK or southern Germany. We could view this as an illustration between 16s and shotgun reports.
It is interesting to note that this seems closer to Family Tree DNA results shown above.
In some cases, it is not where the ancestors came from BUT where ancestor siblings settled, as in Iceland and the Shetland Island. The same can happen with bacteria identification. All that we know is that some components are shared.
Some people get concerned about finding unclassified stuff in their microbiome sample. This does not happen with some labs because they elected not to report what is not classified. Why? It leads to support calls asking for explanations (which means $$$$ spent for the company).
Labs could create synthetic proprietary names for the unclassified to make the issue disappear — that causes the issue to disappear but really does not help.
Bacteria is like the population of a country or the world. They interbred to some extent
Genetic exchanges among bacteria occur by several mechanisms. In transformation, the recipient bacterium takes up extracellular donor DNA. In transduction, donor DNA packaged in a bacteriophage infects the recipient bacterium. In conjugation, the donor bacterium transfers DNA to the recipient by mating.
So doing a microbiome test is like collecting the DNA from a bunch of people at a major airport and then asking: Which country did this person originated from according to their DNA. You will find some people that are good matches to an ethnic origin and some people that are “Heinz 57” aka “Mutts”. These people are unclassified — just like some living components are unclassified.
If I was picked, the answer is pretty clear for the “genus” that fits me
Which agrees with paper records back to 1500
If we want to get a more precise name, the “species”, then some fuzziness appears
There is a possible misidentification, my father’s ancestors (all lines) records go back to 1600 on an island that is low probability.
Lolland/Falster
The process of giving names to bacteria in the microbiome is extremely similar. For some bacteria (like me) we fit into a nice box. For the person below, the box is not so clear
Would this person be classified as Irish or English
Bottom Line
A wise man knows what he does not know, and what cannot be known.
COVID in February 2021. 37 y.o. Male at the time, athletic/fit. Crossfit x 3 a week, playing football weekly Only mild gastritis prior to Covid. No other health issues.
Moderate severity Covid, lots of symptoms.
And then Long COVID and CFS/ME type of symptoms mostly fatigue, PEM and GI problems (pain, food intolerance, bloating..etc) I’d say it’s a moderate/mild case of CFS/ME. But after 18 months still not back to previous levels, can’t walk too long otherwise i crash. I’d say i am around %75.
We have two samples available, one early in Long COVID and one more recent
2021-10-01
2022-08-17
With this type of information, let us start by comparing them. We are fortunate that both samples are similar read quality which reduces fuzziness. Unfortunately, it appears that the microbiome dysfunction has increased in many aspects. One aspect that it has improved in terms of bacteria with very low counts. We went from 74% of bacteria with low counts down to 51% with low count. Ideally we would love to see the low count to drop to a modelled 15%.
I should note that the increase in some Outside Ranges is likely because many of the ranges are 0 to some amount, hence the older sample had less because it was full of different trace amounts. The same apply to many other criteria, the low abundance of many bacteria skewed the criteria to appear better.
Criteria
Current Sample
Old Sample
Lab Read Quality
5.8
5.9
Bacteria Reported By Lab
393
399
Bacteria Over 99%ile
8
1
Bacteria Over 95%ile
26
21
Bacteria Over 90%ile
42
35
Bacteria Under 10%ile
135
160
Bacteria Under 5%ile
56
108
Bacteria Under 1%ile
9
29
Lab: BiomeSight
Rarely Seen 1%
1
0
Rarely Seen 5%
9
8
Pathogens
30
33
Outside Range from JasonH
10
10
Outside Range from Medivere
15
15
Outside Range from Metagenomics
10
10
Outside Range from MyBioma
8
8
Outside Range from Nirvana/CosmosId
22
22
Outside Range from XenoGene
34
34
Outside Lab Range (+/- 1.96SD)
16
8
Outside Box-Plot-Whiskers
49
43
Outside Kaltoft-Moldrup
147
108
Condition Est. Over 99%ile
18
0
Condition Est. Over 95%ile
38
0
Condition Est. Over 90%ile
50
0
Enzymes Over 99%ile
561
115
Enzymes Over 95%ile
793
282
Enzymes Over 90%ile
866
680
Enzymes Under 10%ile
289
294
Enzymes Under 5%ile
149
149
Enzymes Under 1%ile
29
23
Compounds Over 99%ile
480
50
Compounds Over 95%ile
600
214
Compounds Over 90%ile
677
298
Compounds Under 10%ile
581
315
Compounds Under 5%ile
563
242
Compounds Under 1%ile
548
91
I next went to the Krona Charts to try to understand the shifts better. We see a massive increase of unclassified Bacteroides
2021-10-01 Sample2022-08-17 Sample
Going to sample comparison, we see that (genus) Bacteroides was at the 99%ile on both samples. Looking at members of this genus, we see several of the identified species at high levels
Then create a handpicked bacteria focused on Bacteroides and the high species under it. To express another way: Feed the weak and destitute, Bring down the mighty.
Doing the first step, we see at the top of the consensus:
The hand picked collection is below with percentiles
genus Bacteroides 99
species Bacteroides faecis 93
species Bacteroides graminisolvens 86
species Bacteroides ovatus 99
species Bacteroides rodentium 95
species Bacteroides stercorirosoris 91
species Bacteroides thetaiotaomicron 93
species Bacteroides uniformis 97
species Bacteroides xylanisolvens 100
The suggestions for just these are shown below. The pattern is similar to other peoples suggestions with ME/CFS – lots of specific B-Vitamins, dark chocolate etc:
whole-grain barley
sucralose
Caffeine
Hesperidin (polyphenol)
polymannuronic acid
momordia charantia(bitter melon, karela, balsam pear, or bitter gourd)
Bottom line for probiotics to try (add just one new one a week so you can see the response to each). See Simple Suggestions download below for suggested dosages or look them up on 📏🍽️ Dosages for Supplements. Using too small (almost homeopathic) dosages is a common error – the dosages on bottles are determined for profit margin (repeat business) and not from effective dosages from clinical studies.
I would suggest getting another sample 6 weeks after implementing the above to see what the progress is.
As always, review with your medical professional before implementing.
Bottom Line
This patient history and their microbiome are in agreement. The antibiotics suggestions (off label usage) matches the history.
Postscript – and Reminder
I am not a licensed medical professional and there are strict laws where I live about “appearing to practice medicine”. I am safe when it is “academic models” and I keep to the language of science, especially statistics. I am not safe when the explanations have possible overtones of advising a patient instead of presenting data to be evaluated by a medical professional before implementing.
I cannot tell people what they should take or not take. I can inform people items that have better odds of improving their microbiome as a results on numeric calculations. I am a trained experienced statistician with appropriate degrees and professional memberships. All suggestions should be reviewed by your medical professional before starting.
The answers above describe my logic and thinking and is not intended to give advice to this person or any one. Always review with your knowledgeable medical professional.
On Microbiome Prescription we have Look up a modifier of bacteria with many entries. It would be good for everyone if we increase the number of entries especially for atypical items.
YOU can help make it happen!
Find a herb, spice, food, drug of special interest to you
Look at the summary or the full study if available
The study should be on a human or an animal (there is not enough studies to be picky), skip fishes.
Look for bacteria names. We need to know the name and how they shift. i.e.
A Ruminococcus ASV, a Lachnospiraceae Anerostipes ASV, two Lachnospiraceae NK4A136 group ASVs, and a Muribaculaceae ASV were enriched in the vitamin K deficient group, whereas a Bacteroides ASV was enriched in the MK4 group, and a Lactobacillus ASV in the MK9 group.
Then comes the thinking part. Trying to describe the results.
Can you write a sentence such as “Vitamin K may increase Bacteroides and Lactobacillus and a reduce Ruminococcus, Anerostipes ,Muribaculaceae ”
If so, include that in the file sent. (I will verify and it will serve as a double check the reading)
If the study was done on a person (or mouse) with a specific condition, we still include it.
Sometimes you will find that a substance with a common name may have multiple breakdowns.
Ideally, find other names of the substance and search each one.
Sending the Information to Me
I would suggest putting the links (i.e. like https://pubmed.ncbi.nlm.nih.gov/36471554/ ) in a text file or excel and sending to me. Where practical, include a sentence on the impact.
The information sent (when added to database) is available to everyone! This is citizen science! As I remarked to someone earlier today “I do not have a business model, I have a pro bono model“.
I would suggest putting the links (i.e. like https://pubmed.ncbi.nlm.nih.gov/36471554/ ) in a text file or excel and sending to me.
The information sent (when added to database) is available to everyone! This is citizen science! As I remarked to someone earlier today “I do not have a business model, I have a pro bono model“.
Clicking on the book stack will list the studies (and provide links!)
The encoded data can be used to evaluate yours and others microbiome against the patterns reported.
Click Taxon to get the list. Or Candidates to get an abstract list of recommendations without a sample
That is it!!! You spend the time so others do not have to and can act on their challenges with better information!
Email the lists to Research@microbiomeprescription.com
Have you ever heard from anyone who felt significantly worse on antibiotics? Since 2020 when I had ebv and iv antibiotics for an infection I can’t tolerate antibiotics like I used to. Low dose doxy 50mg makes me feel very unwell within a few hours, clarithromycin is better tolerated but if I take that I get severe abdominal pain after meals and reduced physical function which only pro and prebiotics fix. If I take a broad spectrum like metronidazol that completely fucks me up (pardon my french!) and I have to get into bed as I cannot stay awake.
There can be multiple causes, the most likely is a Jarisch–Herxheimer reaction a.k.a. “die-off”. (see this article). I have written about these for many years with a few prior posts.
There are other possible causes, for example, something is being feed and more chemicals are being dumped into your system. That is best determines with lab tests.
So to see explore possibilities (I deal with probabilities, not certainty), For your sample go to My Profile. You will see the new link there under Special Reports. Other common possible causes are probiotics that produce d-lactic acid (often causing brain-fog) or histamines.
Enter what causes you to feel worse on the left size, click compute and see which bacteria are likely being lowered.
A friend got this diagnosis from a naturopath examining blood cells using a microscope. Visual inspection of blood lacks the degree of objectivity that I would prefer (AI imaging of the microscope slides is what I would like to see things evolve to). I went to Microbiome Prescription’s Medical Conditions with Microbiome Shifts from US National Library of Medicine and then look at the candidate suggestions for Nonalcoholic Fatty Liver Disease (nafld) Nonalcoholic .
Out of curiosity, I did a few cross validations for the highest to take and highest to avoid. Everything was in agreement. The few suggested items that I checked were shown to have the desired impact on NAFLD.
The purpose of cross validation is to see how well the Artificial Intelligence Logic (and assumptions) are performing.
I decided to do a deeper cross validation because I found that the treatment literature was relatively abundant. REMEMBER the AI only knows the bacteria shifts and nothing about the diagnosis or treatments.
Items to Take
In priority order, the top items with AI Weight of 20 or more
There is a strong bias to publish positive results, hence finding many n/a in the to avoid list is expected. A study show an adverse effect is unlikely to see publication. We see the following ratios:
To Take: 22 Right to 2 Wrong, i.e. 92% correct
To Avoid: 10 Right to 2 Wrong i.e. 83% correct
This suggests that the n/a ones above are likely to be correct.
A reader forwarded his results and ask if they could be uploaded. It was a CSV file which was a good sign. Inspecting the file I noticed two things:
No NCBI Taxonomy numbers were included 🙁
The report gave percentile numbers for every bacteria — a wonderful thing to see.
The reader approached Thorne support about getting the NCBI Taxonomy numbers added — with no success. After a few days of work I ended up with 99.9% success of matching their bacteria names to NCBI Taxonomy numbers. The import worked… but wait! The price is about the same as some 16s tests, but you get MORE DATA and more accurate data! See this study for the difference between 16s and Shotgun
This is whole-genome shotgun metagenomics which is more accurate. It provides percentiles against a much larger sample than I could hope to get. My site is focused on percentiles — so most thing flows nicely – even when there is just one sample!!!
There are items that will not work until we hit 100+ samples from Thorne (i.e. KEGG Percentile Ranking, Pub Med Condition Percentile, etc).
We use Thorne’s PercentilesInsufficient Data To Use these options
We substitute Percentage Match for Percentile in this section (since we are less than 100 samples)
KEGG is based on Percentiles
Bottom Line
I have ordered Thorne for my next test and expect to keep using them if the pricing stays the same. These test costs are driven by technology — which keep dropping cost over time. I recall spending $1000 to get 1 Meg of RAM many decades ago, today for the same amount, I can get 320 GB of memory — that’s 320,000 x more! The same thing is happening with microbiome and DNA testing technologies.
Recent Comments