Decoding personalized nutritional, microbiome and host patterns impacting clinical and prognostic features in Crohn*s disease

Crohn's disease (CD) presents during childhood in 10-20% of cases and manifests
in chronic relapsing debilitating symptoms. Compared with adults, pediatric CD
is more extensive and aggressive. It is believed to arise in genetically
susceptible individuals via excessive intestinal immune-activation. The factors
responsible for this uncontrolled immune-mediated inflammation are only
partially understood, but perturbations of the gut microbiome are believed to
be critically important. While the intestinal microbiota is specific to each
individual and remains stable for long periods of time, systematic shifts in
its composition and function have been observed in patients with CD, compared
with healthy individuals1. Dietary and nutritional shifts have been
convincingly shown to impact the composition and function of the microbiome.
However a clear actionable link between nutrition, gut microbiome composition
and function, features related to clinical manifestations and the severity of
CD has not been comprehensively investigated. Of note, CD is often responsive
to dietary intervention, namely exclusive enteral nutrition (EEN), which is
considered the first-line therapy in pediatric CD flares2. EEN seems to be less
efficacious in adults compared to children3. Although the mechanism of this
response is unclear, changes in gut microbiota seem to parallel the clinical
response4.
We have recently shown in the largest human cohort to date5,6 that utilizing
advanced computational pipelines, such as machine learning techniques, enables
us to correlate personalized dietary habits, the gut microbiome and
individualized host outcomes, to post-prandial glycemic responses. Moreover,
interventional trials utilizing these person-specific algorithms enabled to
tailor unexpected dietary interventions that normalized glucose levels in
pre-diabetic individuals, providing a proof of concept for the utility of
unbiased integration of *big data* in reaching translational clinical
applications.
In this large-scale multi-national study, we propose to utilize similar
approaches to study nutritional responses in a cohort of newly diagnosed
pediatric CD patients, with an aim to reach new levels of understanding on
features related to individualized inflammatory and metabolic responses of CD
patients to nutritional compounds. Moreover, we intend to collect multi-omic
datasets to devise patient and/or patient subset-specific machine learning
algorithms, enabling to individually employ defined and measurable nutritional
interventions with an aim to integrate them into the therapeutic scheme as
means of improving inflammatory and metabolic profiles in CD patients.
A key success criterion is to evaluate the likelihood to predict, based on
modeling, why disease activity responds to a given dietary intervention in some
individuals, whereas in others it does not. As such, we will (1) develop a
dedicated bioinformatics pipeline that enables primary analysis and
visualization of the data. (2) Correlates metagenomics data with dietary
patterns to quantitatively describe how disease severity parameters respond to
diet. (3) Uses integrative analysis to identify microbial species interactions
and thereby identifies stable consortia of gut symbionts. (4) Develops a
kinetic modelling framework that allows for the simulation of how different gut
symbionts interact with each other, and with host immunity. Specifically, we
will correlate metagenomics data with dietary patterns.
The dynamic nature of our model is capable of identifying nutrients and
subsequent diet-microbiome interactions that may favourably affect CD behavior
and integrating multiple substrates in a complex environment, which is thus
suitable for the investigation of the colon milieu.

By means of this study we want to find out whether there are certain
characteristics that can influence or predict the course of Crohn's disease. We
want to do this by collecting data on diet, inflammation levels in the blood,
complaints and the microbiome of children who have recently been diagnosed with
Crohn's disease. With the use of artificial intelligence techniques, computers
will be able to establish links between dietary patterns and inflammation
levels in patients with Crohn's disease.

This is an international multi-center 3 arm study. The study will include 250
newly diagnosed pediatric CD patients. All of the 250 CD patients will be
recruited in parallel. However part of the data collected will be used for the
primary construction of the algorithmic setup while the other part will be used
for corroboration of the personalized algorithms. We will also recruit 20
healthy controls, undergoing colonoscopy for non-specific abdominal pain and 30
non-invasively characterized healthy controls to enable the machine learning
process to differentiate between normal and pathological signals.
Patients will be allocated, recruited and followed at several leading pediatric
IBD centers around the world headed by the project leads.

The risk involved with drawing blood is minimal, and involves only mild
discomfort. There are well-known risks associated with endoscopy, due to
insertion and maneuvering of the endoscope. However, endoscopies will only be
performed for medical indications and per recommendation of the patients'
physician. No endoscopies will be performed solely for research purposes. We
estimate the risk of additional biopsies for this study to be negligible, in
light of the large body of research that has found no increased risk of
significant bleeding or perforation. In addition, intestinal biopsies do not
cause any discomfort or pain.
No risks to participants are involved in taking samples of stool.