Thomas B. Knudsen, PhDDevelopmental Systems Biologist, US EPA
National Center for Computational ToxicologyCSS - Virtual Tissue Models Project
[email protected] 0000-0002-5036-596x
Computational Systems Toxicology:
Recapitulating the logistical dynamics of cellular response networks
in virtual tissue models
“Advancing Computational and Systems Toxicology for the effective design of safer chemical and pharmaceutical products”EUROTOX 2017 - Bratislava
DISCLAIMER: The views expressed are those of the presenter and do not necessarily reflect Agency policy.
In a nutshell …
• Advances in biomedical, engineering, and computational sciences enable HTS profiling of the chemical landscape (ToxCast/Tox21).
• HTS data streams can support integrated approaches to testing and assessment but must be tied in some way to biological understanding (MOAs, AOPs).
• Considerable mechanistic knowledge exists about cellular networks that pattern tissue development (cell signaling).
• Information must be collected, organized, and assimilated into in silico models that link HTS data (in vitro) to apical outcome (in vivo) and back (predictive toxicology).
2
Virtual Embryo: an array of systems models to forward- and reverse-engineer
developmental toxicity for mechanistic understanding and predictive toxicology.
3
4
• Modeling biological systems is a major task of systems biology, as most cellular phenomena are governed by interconnected dynamical networks.- cell growth, proliferation, adhesion, differentiation, polarization, motility, apoptosis, …- ECM synthesis, reaction-diffusion gradients, clocks, mechanical boundaries, fluid flow, …
• ABMs recapitulate cellular networks show how complex processes are regulated and how their disruption contributes to disease at a higher level of biological organization.- reconstruct development cell-by-cell, interaction-by-interaction- pathogenesis following synthetic knockdown (cybermorphs)- introduce ToxCast lesions into a computer simulation
- return quantitative predictions of where, when and how the defect arises.
Cellular Agent-Based Models (ABMs)
1. Reverse-engineering the system: top-down scaling
• Suppose we know an apical outcome (eg, cleft palate), how far can an ABM take us to inferring a key event?
Hutson et al. (2017) Chem Res Toxicol
SEM
s o
f h
um
an
pa
late
by
K S
ulik
, UN
C
Palatal fusion in silicoPalatal fusion in vivo
5
6
Hacking the control network ‘Cybermorphs’
6
Cybermorph ToxCast lesion: Captan-induced cleft palate in rabbits
Ass
ay r
esp
on
se
EGF
TGFb
µM concentration
fusion no fusion
OUTPUT: tipping point mapped toHTS concentration response
(4 µM)
Captan in ToxRefDBNOAEL = 10 mg/kg/dayLOAEL = 30 mg/kg/day
OUTPUT: tipping point predicted bycomputational dynamics
(hysteresis switch)
HTTK pregnancy model predicts 2.39 mg/kg/day Captan would achieve a
steady state concentration of 4 µM in the fetal plasma
INPUT: Captan in ToxCast
7
2. Forward-engineering the system: bottom-up scaling
• Suppose we know a molecular effect (eg, ToxCast lesion), how far can an ABM take us to hypothesizing an apical outcome?
Saili et al. (2017) manuscript in preparation
BBB Phylogeny BBB Ontogeny - >90 genes, >5 cell types
Mancozeb
8
F Ginhoux, Aymeric Silvan – A*STAR, Singapore
Computational dynamics of brain angiogenesis
Tata et al. (2015) Mechanism Devel
VEGF-A gradient: NPCs in subventricular zone
normal mouse, E13.5 microglia-depleted
We are building and testing computer models formulated around novel hypotheses such as ‘chemical
disruption of microglia perturbs brain angiogenesis’.
9
In silico cascading dose scenario
CSF1RVEGFR3VEGFR2
Mancozeb in ToxCast
INPUT 0.03 µMOUTPUT: predicted dNEL
INPUT 0.3 µM: AC50 CSF1ROUTPUT: fewer microglia drawn to EC-tip cells
INPUT 2.0 µM: AC80 CSF1R + AC50 VEGFR3OUTPUT: overgrowth of EC-stalk cells
INPUT 6.0 µM: AC95 CSF1R + AC85 VEGFR3 + AC50 VEGFR2OUTPUT: loss of directional sprouting
endothelial tip cellendothelial stalk cellmicroglial cell
Zirlinden et al. (2017) manuscript in preparation10
SYSTOX
HTS
HTK
SAR
MPS
AOP
ABM
Computational synthesis and integration
computationalchemistry
bioactivity profiles
kinetics &dosimetry
microphysiological systems
pathways& networks
computationaldynamics
11
Todd Zurlinden – NCCTKate Saili – NCCTRichard Judson - NCCTNancy Baker – Leidos / NCCTRichard Spencer – ARA / EMVLShane Hutson – Vanderbilt U
Special Thanks
12