Towards semi-mechanistic models of disease progression: A translational systems biology approach Ioannis (Yannis) P. Androulakis Biomedical Engineering and Chemical

1
Towards semi-mechanistic models of disease
progression A translational systems biology
approachIoannis (Yannis) P. AndroulakisBiomedic
al Engineering and Chemical Biochemical
Engineering, Rutgers UniversityDepartment of
Surgery, UMDNJ-RWJ Medical School
androulakis ip
NIH Funded
2
From phenomenological relationships to
semi-mechanistic representations

• (TSB) Mathematical formalisms which use
  mechanistic information and basic knowledge in
  order to simulate behaviours at the organism
  level providing a mechanistic basis for
  pathophysiology
• Link outcomes (clinical responses) to processes
  (cellular mechanisms)
• What should be monitored (state
  variables/markers)
• How are the states variables interconnected
  (network topology)
• How can the dynamics be inferred (progression
  dynamics)
• From causal relationships to progression models
  through vignettes in
• Interpreting data
• Interpreting models

3
Hierarchy of models, data and methods
Models (experimental)
Data Biochemical (cellular, tissue and host
level) Outcome (physiological at the systemic
level)
Models (computational)
Phenomenological models
(semi)-mechanistic models
Knowledge representation models
Methods (computational)
Statistical correlations
PLSR
Bayesian models
ODE models
4
Context dependent causal relationshipsTissue-spe
cific gene expression
5
Context dependent causal relationshipsCirculating
cytokines
6
Integration across layers of informationGene
expression, signalling, cytokines
7
Dynamics A step beyond causal relationshipsFrom
genes
8
Dynamics A step beyond causal relationshipsto
clusters
9
Dynamics A step beyond causal relationshipsto
functions .
10
Dynamics A step beyond causal relationshipsPoint
s of divergence from homeostasis
While the system diverges from homeostasis, not
everybody diverges at the same time at the same
rate in the same direction
11
(disease) Progression as evolution along dynamic
trajectories
Homeostatic regulatory mechanisms adequately
control certain systemic responses
Homeostatic training with limited data
12
Disease progression as evolution along dynamic
trajectories
13
Disease progression as evolution along dynamic
trajectories
State variables are close, yet outcomes differ
substantially
14
Disease progression as evolution along dynamic
trajectories
poor mans clinical trial Parameter sampling
simulates population heterogeneity Same
initiating response induces responses which
live on well defined manifolds
15
Disease progression as evolution along dynamic
trajectories
Markers reproducing systemic dynamics
16
Intrinsic computational model dynamics precede
(predicted) biomarker changes
Can computational prediction of critical
bifurcation events precede (and inform)
outcome? The biomarker becomes an interpretation
of the dynamics
Can computational prediction of critical
bifurcation guide evaluation?
17
A multi-scale translational systems biology model
of human endotoxmiea
18
RR(t1)
RR(t1)
RR(t)
RR(t)
19
Holzheimer et al., SHOCK 2002
20
(No Transcript)
21
Suppressed rhythms and the HPA axis
Disrupted chronobiology has profound implications
in the context of (i) indicator of disease state
(ii) instigator of disease (iii) compromising
factor (chronic stress)
22

• Semi-mechanistic computational models can,
  potentially, enable
• the integration of information across multiple
  levels (cellular, systemic, physiologic)
• the in silico assessment of heterogeneity

23
The digital patient (T. Buchman, MD)

• Mathematical models
• can establish links across multiple scales (i.e.,
  genomic, proteomic, metabolomic)
• encourage the subtle distinction between outcomes
  (clinical response) and processes (cellular
  mechanisms). The latter are most likely
  responsible for individualized responses
• Computational model that capture mechanistic
  information, at a reasonable level
• can help establish complex markers characteristic
  of disease progression
• can form the basis of in silico clinical trials
  reflecting heterogeneities at various levels
• Mood disorders and cytokines, loss of rhythms and
  circadian disruption, sleep loss, stress
  (physical and psychological) ?
  Patient history