A fundamental goal of Microbiome Prescription is not to be prescriptive (i.e. this is the right way of doing it) but supportive of many ways that readers have requested. There are many approaches that I suspect are very poor choices — but that is an individual’s choice. A few examples are below. Most are based on percentages and a few on percentiles. In addition to these canned choices, users may individually select the bacteria and get suggestions from their choices.

“Just give me suggestions”

This option arose at the request of people with cognitive impairment, limited education and brain fog. It is executing what I deem to be the best algorithms that are then fed into a Monte Carlo model (aka CONSENSUS) to get suggestions with likely very good odds of being correct. I use “odds” because every thing is based on estimated probabilities, hence the fuzzy logic expert system description.

Recent work and research has lead me to modify the pick of algorithms, in the hope of improving the odds more.

Algorithms

• Step 1 is always to remove existing suggestions. You can add those back in later if desired.
• Step 2 is selecting bacteria that are out of range by the canned standard lab formula [Shifts_Lab]
  • i.e. outside of Mean +/- 1.96 Standard Deviation
  • Usually the number of bacteria selected is low
• Step 3 is selecting bacteria that are out of range by the Box-Plot-Whisker ranges
  • This usually picks more bacteria than above
• Step 4 is selecting bacteria that are out of range using the patent-pending Kaltoft-Moldrup algorithm
  • This picks bacteria whose values are “visually out-of-whack”
• Step 5 is working only for matches to the top statistically significant bacteria shifts based on statistical associations to symptoms (see Technical Note: Identifying Key Bacteria to Address for more information)

Some bacteria will be picked in all of the above, others in just one. Picking the bacteria is not the focus, it is picking the substances that will correct them based on the computations of the expert system. The same substance may be picked for each different collections of bacteria above. It is the common substances that is important.

“Just give me suggestions with Symptoms”

Does all of the above, and then include any bacteria with shifts matching the shifts of other people using the same lab that reported the same symptom. Conceptually, it will give a boost to the most important bacteria for a person’s suggestions

Example of Monte Carlo model / Consensus

The weightings (Priority/Take Net/Avoid Net) are proprietary. The facts (some 1.3 millions) applied are shown for those that wish to devise their own weights. The facts are shown by clicking the 📚 icon. An example is shown below.

Example of the facts used:

Some facts are favorable and others unfavorable. As with a human expert, the fuzzy logic expert system makes inferences. For example, inferring that Lactobacillus casei would have a similar effect as Lactobacillus casei Zhang. Similarly, there will be some (reduced) impact on the parent and children of what is reported in a study. For clarity, something that impacts a species will be inferred to have a reduced impact on the genus and a reduced impact on all of the strains.

Some examples of undesired impacts.

Bottom Line

The goal of Microbiome Prescription is to give choices with automation of what is believed to be the best approach. If someone is not happy with the choices, they can do what they want by hand using the information provided. Percentages and Percentiles are shown for all samples. The facts used to make decisions are shown.