The metabolic potential
of
microbial communities
• Maria Persico, PhD
• Postdoc, Integrative Bioinformatics and Biostatistics
maria.persico@recetox.muni.cz
Microbiome Metabolome Integration
Why this research line?
Microbiome Metabolome Integration Platform (MMIP): a web-based platform for
microbiome and metabolome data integration and feature identification., https://doi.org/10.1101/2023.04.04.535534
The Invisible Us – The Human Microbiome in Health and Disease, https://dx.doi.org/10.31487/sr.blog.07
Microbiome
• Microbiome - a community of microorganisms that can usually be found
living together in a given environment
• Microorganism - a single-cell organism of microscopic size
• Bacteria
• Viruses
• Fungi (Brewer's yeast is a eukaryote belonging to this kingdom )
• Algae
http://www.wikiskripta.eu
http://aboutviruses.weebly.com
The human microbiome
• We have more bacteria in our body than
our own cells
•      Bacterial genes outnumber human
genes 100:1 
•      More than 1000 species of bacteria
live in the intestine
•      Based on the microbiome, a person
can be identified in a similar way to
fingerprints 
•      Each person has an individual
composition of the intestinal
microbiome, it differs from 80-90% 
• Gut microbiome and discovery of Gut
Brain axis: the two-way biochemical
signaling that takes place between the
gastrointestinal tract (GI tract) and our
central nervous system (CNS)
The Invisible of us https://dx.doi.org/10.31487/sr.blog.07
Microbiome in healthy status:
contributing internal and external factors
Microbiota in health and diseases, https://www.nature.com/articles/s41392-022-00974-4
How does the microbiome affect health? 
Gut
microbiome
Digestion of
food
Synthesis of vitamínes
(K a B)
Prevention of the
multiplication of
pathogenic bacteria in
the intestines
Stimulation of immune 
systém
"Training" of the immune
system after birth
Effect on fat storageProper development of
the intestinal wall
Influence on brain
development
When something doesn't work...
Microbiota in health and diseases, https://www.nature.com/articles/s41392-022-00974-4
• Dysbiosis, a state of disruption of the balance of the
microbiome and resulting changes in its composition
and function
How does the
microbiome affect health?
• By its metabolic activity:  
• it is processing something 
• It creates something
Metabolic control by the microbiome | Genome Medicine | Full Text (biomedcentral.com)
Microbiome research in health asks 3 basic questions
• Simple - we find out what bacteria are in the gut -> we
make a genotype -> we estimate the functions
How to answer these questions? -”Who”
•      Simple - we find out what bacteria are in the gut -> we make a
genotype -> we estimate the functions
•      Traditional procedures - cultivation? 
•   The problem: most bacteria in the gut are not culturable
How to answer these questions?
Metagenomics
• Study of the genomes of all microorganisms in the sample (soil,
water, skin smear, feces, tumor...)
GENOMES
BACTERIA FUNGI
How to explore the metagenome?
Marker metagenomics
(targeted sequencing)
Amplicons corresponding to the whole (or
parts) of genes of so-called phylogenetic
markers (16S rRNA, rpoB...) are isolated,
extracted and sequenced.
Marker genes are used as “speciesspecific
taxonomic barcodes” – a rapid
estimate of taxonomic composition
How to explore the metagenome?
Shotgun metagenomics
(whole genome
sequencing) 
Marker metagenomics
(targeted sequencing)
The entire genome of the microbiome
in the sample is extracted and
sequenced.   It provides insight into
the taxonomic composition and
function of the microbiome. 
Amplicons corresponding to the whole (or
parts) of genes of so-called phylogenetic
markers (16S rRNA, rpoB...) are isolated,
extracted and sequenced.
Marker genes are used as “speciesspecific
taxonomic barcodes” – a rapid
estimate of taxonomic composition
Marker metagenomics
(targeted sequencing)
Taxonomic
composition
Shotgun metagenomics
 (whole genome sequencing) 
Proteomes and functional annotations of proteins
How do we get answers for “what” and
“How”.?
We find out the function of genes
from web knowledge base
(knowledgebases) about genes,
their functional products and their
involvement in molecular
pathways
application of special bioinformatics
tools:
PICRUST + PRMT, METAPHLAN,
….
KEGG, REACTOME, UNIPROT, ...
By sequencing we
will find out what the
microbiome is and
its genome
We have information about the
composition and functional
POTENTIAL of the microbiome
(metabolic pathways and
metabolites) 
How to bring a sick person closer to a healthy person?
 Hypothesis: the microbiome affects health through metabolites => changing the
microbiome of a sick person can help to change his metabolites and thus his
health status
?
But what do we not know
(gap of knowledge)?
• How to use a list of differently abundant bacteria or differently
expressed metabolites to treat a patient - to change their
individual microbiome...
A typical data integration strategy in
Microbiome Metabolome Integration
 Metabolic and
microbial profiles of
individuals
Comparison
between groups
Healthy population
and diseased
population
List of different metabolic
pathways (Picrust)
List of different metabolites
(PRMT method)
List of differences in
microbiome composition
Microbial composition
Microbial composition
What do we need? 
A method that estimates the microbial composition based on the desired (or target) metabolic profile
The method can be implemented to end up with a software tool.
Microbial composition
Target metabolic profile
Start metabolic profile
What is the target metabolic profile
of a healthy individual?
Thanks to our clever experimental
design we have also collected data
from healthy individuals...
We have microbial composition of healthy people. What do we need?
A method that estimates the microbial composition based on the desired metabolic profile
Microbial composition
Microbial composition
Using an adaptation of the PRMT method we can derive a
Microbial dictionary of biochemical/metabolic
functions, containing functional/metabolic information
about all microbes in all individuals we collected data from
( patient specific microbial composition and healthy
individual microbial compositions).
which microbes
produce which
metabolites?
Metabolity Skóre
M1 50
M2 23
M3 -5
M4 8
Metabolity
A type of microbe M1 M2 M
3
M4
Methanobrevibacter 2 0 1
0
5
Oscllibacter 5 0 3 0
Akkermansia 0 4 1
2
0
Treponema 1 6 0 0
We do not know
yet how to derive
the desired
metabolic profile
( reference
estimated
metabolome
...see next slide :)
"Treponema"=11.02​
 "Oscillobacter"=3.05​
"Bacillus S."=7.18​
 "Clebsiells" =1.29​
…
…
...
"E.Coli" =6.77​
Larsen PE, Collart FR, Field D, Meyer F, Keegan KP, Henry CS, McGrath J Quinn J, Gilbert JA Predicted Relative Metabolomic Turnover (PRMT):
determining metabolic turnover from a coastal marine metagenomic dataset. Microb Inform Exp. 2011 Jun 14;1(1):4. doi: 10.1186/2042-5783-1-4. 
.
What do we need?A proper mathematical formulation for the problem
Microbial composition
Microbial composition
W
PRMT metabolites
scores associated to
microbes in our
dictionary
Metabolity PRMT
score
M1 50
M2 23
M3 -5
M4 8
A type of microbe %
Methanobrevibacter ?
Oscllibacter ?
Akkermansia ?
Treponema ?
X =
Thanks to the
matematical
formulation we
gave to the
problem, we can
derive the
metabolic profile
of both healthy
and diseased
individuals
W
Metabolity
Druh mikrobu M1 M2 M
3
M4
Methanobrevibacter 2 0 1
0
5
Oscllibacter 5 0 3 0
Akkermansia 0 4 1
2
0
Treponema 1 6 0 0
"Treponema"=11.02​
 "Oscillobacter"=3.05​
"Bacillus S."=7.18​
 "Clebsiells" =1.29​
…
…
...
"E.Coli" =6.77​
"C00084"=11.02
 "C00473" =-3.05
"C05577"=-7.18
 "C07490" =1.29
"C16551"=-1.70
"C16587"=3.01
"C16596"=4.47
"C00577"=10.00
"C00116" =-6.77
 "C00441"=25.00
"C00191"=0.4
"C11402"=21.46
"C04517"=-87.46
-"C03752"=31.02
"C06192"=8.77
"C11418"=-750
d
d
W
d
=
S
XMicrobial
dictionary
of
biochemical
/metabolic
functions
MiMetDec – a method of deconvolution of microbial profiles based
on their metabolic potential
Principle: Basis pursuit functional approximation
What does it do?The tool takes an (estimated) metabolic profile, a library of microbial profiles and estimates the
microbial composition that would lead to that d reference metabolic profile
The output of the procedure is the w^ or rebalanced microbial community.
How can we use this method?
Find all microbial compositions capable of providing the same metabolic profile (phenotype).
To find out how to specifically modify the microbial composition of the environment (e.g. intestinal microbiome) to obtain the
desired metabolic profile.
To find out which microbes are most important for a certain type of metabolism
A type of
microbe​
%​
Methanobrevib
acter​
?​
Oscllibacter​ ?​
Metabolity​ PRMT score​
M1​ 50​
M2​ 23​
Metabolity
A type of microbe M1 M2 M
3
M4
Methanobrevibacter 2 0 1
0
5
Oscllibacter 5 0 3 0
Akkermansia 0 4 1
2
0
Treponema 1 6 0 0
What do we need to explore microbes as bioterapheutics?
A method that estimates the microbial composition based on the desired metabolic profile
Microbial composition
Microbial composition
We can explore in
silico intervention
with probiotics or
putative
biotherapeutucs!
we can expand our
intial dictionary of
microbial metabolic
functions
Metabolity PRMT
score
M1 50
M2 23
M3 -5
M4 8
A type of microbe %
Methanobrevibacter ?
Oscllibacter ?
Akkermansia ?
Treponema ?
X =
We have already
derived the
(reference)
metabolic profile of
healthy
individuals...the
closest for each
patient...its 
healthy prototype
Metabolity
Druh mikrobu M1 M2 M
3
M4
Methanobrevibacter 2 0 1
0
5
Oscllibacter 5 0 3 0
Akkermansia 0 4 1
2
0
Treponema 1 6 0 0
Lactobacillus 0 0 1 10
The tool will estimate for us the
composition of the new 
therapeutic microbial community 
What must be the microbial composition
(the abundance of individual microbes of
the patient) in order to achieve the
metabolic profile of a healthy individual?
~
d
How does the tool work in a real example? - the gut microbiome in
colorectal cancer
We exploited already processed data(Wirbel’s
validation cohort:
22 patients with CRC, 16 healthy controls
Metagenome sequencing from fecal samples =>
species composition of bacteria=> our best guest
strain resolution level
 
Methodology:
1. Estimation of the metabolic profile of hypotized
strains 
2. Finding a bacterial healthy prototype (11 found)
3. Estimation of changes in the patient's microbial
profile based on a healthy prototype (d^=Sw^)
The estimated metabolic profiles of
patients and controls suggest
significantly different profiles even in
healthy individuals
Wirbel
et al, Nature. https://doi.org/10.1038/s41591-019-
0406-6
% composition of
the microbiome
In silico experiment,where we have our estimated metabolomes:
1)the original computationally derived metabolome of the patient (black square)
2)the computationally derived metabolome of the healthy prototype (green light)
3)the computationally derived metabolome of the rebalanced microbial community
(that is the output of our method) (dark green)
4)the computationally derived metabolome after a bioterapeutics based intervention (feeding the procedure with
original patient flora + bioterapeutical community)(orange)
How did the patient approximate the metabolic profile of the healthy prototype in the in silico experiment?
Representing the d vectors (alias computationally derived metabolome )
in the space of Principal Components
The closest
healthy prototype
   The most
distant healthy
prototype
The original
patient profile   
Healthy
prototype
After
microbiome
of the
patient
changed
compared
to the
prototype
with the
addition of
probiotics
Example of one patient
Altered
patient
microbio
me
relative
to the
prototype
Healthy prototype
OriginalInPatient
Insilicorebalancing
Insilicorebalancing
withprobiotics
(d^=Sw^)
d vectors
(alias computationally derived
metabolome )
Example of one patient
How did the patient
approximate the metabolic
profile of the healthy
prototype?
Healthy
prorotype 
The original patient
profile
The most
distant healthy
prototype
The closest
healthy
prototype
Altered
patient
microbiome
relative to the
prototype
Altered
microbiome
of the patient
compared to
the prototype
% composition of
the microbiome
Altered
patient
microbiome
relative
to the
prototype
changed
microbiome
of the
patient
compared
to the
prototype
with the
addition of
probiotics
The closest
healthy prototype
The original
patient profile   
The most distant
healthy
prototype   
What's next?
 - implementation of the method into a package in
R/Bioconductor  
- incorporating the effects of XENOBIOTICS
  - incorporating the HOST's metabolism
External collaborator
Daniela de Canditiis​
Italian National Research
Council | CNR · Institute
for Applied Mathematics
"Mauro Picone" IAC 
Thanks for your attention :)
Microbial composition
Microbial composition
What do we need?
A method that estimates the microbial composition based on the desired metabolic profile
What is the metabolic profile of a
healthy individual?
Thanks to our clever experimental
design we have also collected data
from healthy individuals...and by
interrogating the appropriate
database....
IF WE WANT …
 - to identify association between a
microbe and a specific metabolite
produced or consumed by this microbe,
we can perform a kind of leave one out
procedure 
- filtering the result is of fundamental
importance
  -computations are quite efficient
How does the tool work in a real example - the gut microbiome in colorectal cancer –
estimated relative abundances of strains in in silico adjusted profiles
%composition of the
microbiome
Healthy
prototype
Original profile
patient X
Adjusted profile
in patient X
Which bacteria changed most often and their elimination caused
the biggest problems? 
What effect
does the
elimination of
the bacterium
have on the
metabolic
profile?  
The bigger the
number, the
bigger the
influence. 
Microbes
(taxon ids)
Patients
Comparison of the microbiome in the stool and in the tumor
The tumor microbiome is different from
the microbiome of healthy tissue and
stool  
It is enriched with potential oral
pathogens
TMS – tumor microbiome subtypes  
Tumors are divided into three basic
groups according to their microbiome
Zwinsova et al., 2021, Cancers 10.3390/cancers13194799 
How does the tool work in a real example? - the gut microbiome in
colorectal cancer
Colorectal cancer - a very heterogeneous
disease.
Bacteria affect the tumor
 - positively: they expose the tumor to the immune
response   - 
- negatively - they worsen the prognosis, hide the
tumor from the immune system, influence the
response to therapy, cause additional mutations with
their genotoxic products
*"Bacterial passengers of CRC are defined as gut
bacteria that are relatively poor colonizers of a
healthy intestinal tract but have a competitive
advantage in the tumour microenvironment, allowing
Bacterial "driver-passenger" model* of colorectal
cancer development
https://link.springer.com/article/10.1007/s12094-021-02738-y/figures/3
IF WE WANT …
 - to identify a small community of microbial strains having
functional similarity with a microbe leader that we already know
as probiotics, we can build a reference metabolome (d) ad hoc
to pursue this goal  
- using E Coli Niesle as microbe leader, we identified other 13
microbes constituting together a putative therapeutical cocktail
of microbes
  - the tool estimates the relative abundances of each member
of the therapeutical cocktail
Description
of modules
Taxon ids of microbilal strains
selected by the tool when an
ad hoc reference metabolic
profile has been provided as
constraint of the problem