The following looks at a holisitic approach to generate suggestions for microbiome dysfunctions, symptoms (that may be microbiome associated) and diagnosis (that have microbiome patterns).

This model (or variation there of) is being used by several microbiome testing companies today. See the bottom for example of clinical success.

This post illustrate the process and is not a precise match for current implemenation on Microbiome Prescription (which continuously evolves over time).

Native taxa weights

The first step is to get a weight for each taxa in a sample to identify what should be altered and the importance of each. With shotgun samples, there may be over 7000 different taxa.

The simple first step is to just do a lookup compare to ranges for each taxa (assuming there is sufficient data to compute ranges). Then assign weights based on the sample positioning in the ranges. The key function (tax_range) is often a complex function which may incorporate percentage, percentile, gender, age, diet style, and bacteria hierarchy. For example, Lachnospiraceae bacterium GAM79 may dominate and result in Lachnospiraceae being given no weight and thus expert system rules may be involved.

Conceptually, it is the importance of a bacteria to be shifted with the desired direction of shift converted to a numeric value or vector of values.

This is called a native taxa weights .

Presentation taxa weight.

These native taxa weights are then modified by the presence or absences of diagnosis and symptoms. Conditions are not either/or. A good example is Autism which has a wide spectrum of levels. A bacteria known associated with a condition will likely have an increase weight. A bacteria with no known associations will have a decreased or no weight. This is called a presentation taxa weight. As above, it may be a single value or a vector of values.

Modifier Matrix

We drop the taxa weight into our grid as show below. We show the weigh as a single value below. With a positive weight indicating something to increase and a negative weight indicating something to decrease. The “-1 to 1” indicates a factor.

We now want to maximize the value of the suggestions, i.e.

Sum _{Over All Bacteria}( Factor_{Vit B1} * Amount_{Vit B1} +Factor_{Vit B2} * Amount_{Vit B2} + etc)

Amount often becomes a 1 or 0 (take or do not take) when there is no dosage related data. Factor_modifier may be multidimension function on occasion. For example, it values may depend on other factors being selected. This can result in iterations that was the goal the Simula programming language. That is, you get the first naive suggestions(no dependencies), then feed the results into the next iteration.

We can rotate our focus to obtain lists of “to take” and “to avoid”

Sum _{Over All Bacteria}( Factor_{Vit B1} * Amount_{Vit B1})

Factors are often computed from a variety of factors, a few examples:

• the number of studies reporting a shift (often studies disagree),
• the magnitude of the shift (and/or P value),
• the modifier (a specific probiotic strain, a probiotic mixture, a species)
• context of the studies (humans, mice, pigs, fish, fouls).

Then We enter the Casino…

Rather than arguing over exactly which formulae for weights are correct. We make use of multiple reasonable formulae. Each is run independently and we then apply Monte Carlo modelling to these results.

Linearity is Dangerous To Assume

Our experience is that assuming linearity produces poor results. We found that doing cross validation allows this host of functions to be tuned.

Inferences should also be factored in, i.e. if a modifier alters Lactobacillus genus without details on individual species, most people will assume that it will alter some of the species — unfortunately, there are many studies reporting that lactobacillus increased with some species decreasing and other increasing.

The key issue is dealing with very sparse data that is often heavily conditioned, i.e.

• Shift(“Wormwood”, “Bifidobacterium” | Autism) < 0
• Shift(“Wormwood”, “Bifidobacterium” | Diabetes) > 0
• Shift(“Fasting”, BacteriaA | “Chinese” ) > 0 (see Ramadan Fasting Leads to Shifts in Human Gut Microbiota Structured by Dietary Composition – PMC (nih.gov) )
• Shift(“Fasting”, BacteriaA |”Pakistan”) < 0

This may explain why wieghts can be vectors of values.

This is where the art of microbiome manipulation comes in.

Clinical Success

Personal Experiences

Via our free for personal use (not commercial/medical office use) we have had many people have done a sample with one of many supported labs, obtained suggestions from the above model and implemented some, and then done a second sample. For everyone that has done this, there has been OBJECTIVE and SUBJECTIVE improvement. I was expecting > 50% only, but we are running 90+%.  For example analysis from those who consented to share, see this collection dealing with Long COVID and Chronic Fatigue Syndrome.

A recent example is shown below using multiple “measuring sticks” from different labs. We see clear improvement.

We also have associations of symptoms to bacteria using our 5000+ donated samples annotated with symptoms. Often the associations exceed P < 0.001 on a lab specific basis. From this data we can give percentage estimates on pattern matching to symptoms. Below is an example for the person shown above.

We see improvement across all of the top symptoms.

We do not look at “cure” (that does happen sometimes), but reduction of symptoms as our criteria.

We have had incidental reports of it appearing to improve the success rate and speed of remission for some cancers.

AI Cross Validation

Additionally we have done cross validation against the literature.  We take the microbiome shifts reported for a condition across multiple studies, run those shifts through the engine, then see how many of the top suggestions have been found to improve this condition according to published studies using those suggestions.  An example is here: Cross Validation of AI Suggestions for Nonalcoholic Fatty Liver Disease .

While not a clinical study as such, it shows that our suggestions appear to agree with results from third party clinical studies.

Here we hit a philosophy crossroad (and often a zebra crossing/speed bump of medical practitioner ego and/or arrogance).

• The road most travelled is focusing on the bacteria most heard about and trying to address them one by one.
  • It keeps the microbiome simple, naively simple. “All you have to do to raise your lactobacillus and bifidobacterium by taking my preferred probiotic mixture [which I will sell to you].”
  • It ignore the need to keep current on recent studies. Chart below is from PubMed. There are almost 25,000 new studies a year or 68 new studies a day.

• The road that I take is to ignore this chatter, and aim to adjust everything in one pass using mathematical models. No favorites bacteria to focus on (without firm evidence from studies that it is critical for a symptom or diagnosis).
  • I view this approach is most likely to cause desired changes and not chasing this bacteria or that bacteria is isolation.
  • It is accepting microbial interdependence in all of it’s complexity (see below)
  • Using KEGG: Kyoto Encyclopedia of Genes and Genomes data for Metabolites and Enzymes, I do not go down the rabbit hole of some substance being produced by just one bacteria or small set of bacteria. I accept the full width of the microbiome.

Gut Microbiome

The human gut hosts a diverse and complex microbial community:

• Over 10,000 microbial species have been identified in the human ecosystem, with the majority residing in the gut.
• Gut bacteria contribute about 8 million unique protein-coding genes, which is 360 times more than human genes. These bacterial genes are critical for human survival, as they enable us to:
  • Digest foods and absorb nutrients that we cannot process on our own
  • Produce beneficial compounds like vitamins and anti-inflammatories

Microbial Interdependence

Microbial interdependence refers to the complex relationships and interactions between different microorganisms in a community, where they rely on each other for survival and functioning. Here are some key aspects of microbial interdependence.

This study illustrates some interactions, one bacteria reduced a lot of other bacteria. Taking a probiotic that reduces this bacteria, and restore other bacteria.

“[Heyndrickxia coagulans] supplementation improved the gut microbiota imbalance by reversing the decreased numbers [caused by E Coli] of Enterococcus, Clostridium and Lactobacillus in jejunum and Bifidobacterium and Lactobacillus “

Bacillus coagulans prevents the decline in average daily feed intake in young piglets infected with enterotoxigenic Escherichia coli K88 by reducing intestinal injury and regulating the gut microbiota [2023]

Nutrient Sharing

Many microbes cannot produce all the nutrients they need and depend on other microbes to obtain essential compounds:

• The vast majority of microorganisms require nutrients like amino acids and vitamins that they cannot synthesize themselves.
• Corrinoids (vitamin B12 and related compounds) are an important example – while most microbes use corrinoids, only a subset can produce them.

Metabolic Cross-Feeding

Microbes often exchange metabolic products in mutually beneficial relationships:

• Some bacteria break down complex molecules that other species then use as food sources.
• Waste products from one species may serve as nutrients for another.

Symbiotic Relationships

Many microbes form close, interdependent associations with other organisms:

• Corals have symbiotic relationships with algal cells living within them.
• Lichens are symbiotic associations between fungi and algae or cyanobacteria.
• Gut bacteria in animals help digest plant material the host cannot break down alone.

Community Assembly and Function

Microbial interdependence shapes how communities form and operate:

• Public goods sharing drives adaptive function loss and the rise of metabolic cross-feeding over evolutionary time.
• Interdependent patterns that emerge through reductive evolution can make communities more resistant to environmental perturbations.

Ecosystem Roles

Microbial interactions contribute to important ecosystem processes:

• Soil microbes like mycorrhizal fungi and nitrogen-fixing bacteria form symbioses with plant roots.
• Microbial communities in oceans, soil, etc. carry out crucial nutrient cycling.

Understanding these complex webs of microbial interdependence is crucial for fields like ecology, medicine, and biotechnology. It highlights how cooperation and mutualism, not just competition, shape biological communities.

Microbial interdependence occurs when different bacterial species rely on each other for growth or survival. This can happen through various mechanisms:

• Metabolic cross-feeding: One species produces metabolites that another species uses for growth.
• Signaling interactions: Chemical signals from one species trigger responses in another.
• Modification of the environment: One species alters the local environment in ways that benefit another species.

Metabolic Interdependence

• Different bacterial species in the gut perform complementary metabolic functions. For example, some bacteria break down complex molecules that other species then use as food sources.

Colonization and Development

• Infants acquire their initial microbiome from their mother and other caregivers. Even one-day-old pre-term infants have unique microbiomes that differ from each other and their mothers.
• The developing infant microbiome is shaped by factors like genetics, environment, and immune system interactions.

Community Dynamics

• Microbial communities in the human body demonstrate properties like stability (resistance to change) and resilience (ability to return to initial state after perturbation).
• These dynamics can be studied through longitudinal sampling, for example, before, during, and after events like surgery or antibiotic treatment.

Site-Specific Communities

• Different body sites host distinct microbial communities adapted to those environments. For instance, the skin, gut, and mouth each have their own characteristic microbiota.

Examples from Research

Several studies have documented cases where the abundance of one bacterial species depends on the presence or amount of another:

• In the human gut microbiome, researchers have observed that the growth of certain Bacteroides species depends on the presence of specific Ruminococcus species.

Bottom Line

My approach is a holistic approach that attempts to use all of the facts to be considered. At present, over 2.5 million facts or rules. This is based on almost 13,000 studies. The suggestions may not be perfect, but they seem to be both reasonable (strong cross validation is common) and effective for many people.

The alternative paths often is based on “it worked for John Doe, so it should work for you”, or reading a handful of studies (often just one is sufficient for some people to claim being an expert).

When someone tries to “sell you” on their approach ask them:

• How many of the 10,000+ known bacteria do you consider? What is the evidence for excluding bacteria from consideration?
• How many of the thousands of metabolites do you consider? What is the evidence for excluding metabolites from consideration?
• How many studies do you review each month? For myself, it is close to 600 new studies that are identified as worth manual review.

Remember the old analogy of the broad path full of people taking the easy and popular way, versus the narrow path with very few on it.