GRID REQUIREMENTS OF IN SILICO ONCOLOGY:
THE ACGT PROJECT PARADIGM
G. Stamatakos
[email protected]
EGEE’08 CONFERENCE
22-26 Sept. 2008
ACKNOWLEDGMENTS
Special thanks are due to
Dimitra Dionysiou, ICCS-National Technical
University of Athens
Manolis Tsiknakis, FORTH, Greece
Norbert Graf, Dept. Hematology and Oncology,
University of Saarland
and
the whole ACGT Consortium for their
multifaceted contribution to the development of
the Oncosimulator and its positioning within the
overall ACGT architecture
IN SILICO ONCOLOGY
AND THE CONCEPT OF
THE ONCOSIMULATOR
3rd International Advanced Research
Workshop on In Silico Oncology:
Advances and Challenges
September 23- 24, 2008
Zografeio Lyceum, Istanbul, Turkey
http://www.3rd-iarwiso.iccs.ntua.gr/
WHAT IS THE ACGT“ONCOSIMULATOR”
AND WHAT IS ITS PURPOSE ?
The “ACGT ONCOSIMULATOR” is a a
clinically meaningful,
scientifically sound,
technologically advanced and
user friendly
system able to spatiotemporally simulate within
well defined reliability limits tumour growth and
tumour and [to a lesser extent] normal tissue
response to (chemo)therapeutic schemes
for the cases of breast cancer and nephroblastoma
(Wilms’ tumour) in the patient’s individualized
context.
The final goal of the the “Oncosimulator” is
to contribute to the optimization of cancer
therapeutic strategies through conducting in
silico (=on the computer) experiments based
on the individual patient’s multilevel data in
order to support the optimal selection of the
treatment scheme to be administered.
An additional target is to support the design
and interpretation of new clininicogenomic
trials.
SYNOPTIC BLOCK DIAGRAM OF THE ONCOSIMULATOR
THERAPY DECISION
NO
FURTHER
SCHEME?
YES
PREDICTION
EVALUATION
PREDICTION
IMAGING DATA
TUMOR & NORMAL
TISSUE RESPONSE
SIMULATION
CANDIDATE
THERAPEUTIC
SCHEME
RADIOBIOLOGICAL
PHARMACODYNAMIC
PARAMETERS
GENE/PROTEIN
NETWORK
GENE EXPRESSION
DATA
(MICROARRAYS)
GENOTYPING
BIOPSY
MATERIAL
BLOOD
SAMPLE
PURPOSE OF THE DEMONSTRATOR
To show how the “Oncosimulator” i.e. the ACGT in
silico oncology component would look like and
function.
To show how real clinical trials can be simulated in
silico through the integration of basic science
knowledge.
To outline how Grid services are expected to be
utilized by in silico oncology.
To outline how advanced visualization techniques
can be used in order to make predictions easily
understood in the four dimensional context
As the ACGT medical data [imaging,
histopathological, molecular and clinical] are still in
the process of reliable, safe, legal and ethical
distribution within the ACGT consortium,
the “Oncosimulator” concept and construct is
demonstrated using previous work done at ICCS
during the last ten years coupled with new work
done by ICCS, PSNC, UvΑ and other partners within
the framework of ACGT.
CLINICAL CONTEXT OF THE DEMONSTRATOR
Two arms of the Radiation Therapy Oncology Group
(RTOG) Clinical Study 83-02 are simulated in silico.
The study concerns glioblastoma multiforme treated
with radiotherapy [17]
Imaging data [from the Whole Brain Atlas ]
concerning a glioblastoma have been used along with
the radiobiological alpha and beta values that
correspond to a specific glioblastoma cell line with
mutant p53 gene [molecular data] [17]
The exact type of tumour [glioblastoma multiforme]
corresponds to the histopathological data here.
THE TWO RTOG STUDY 83-02 SIMULATED
1) AHF-48Gy:
accelerated hyperfractionation, 48Gy total dose,
(1.6Gy twice daily to a total dose of 48 Gy)
2) HF-81.6Gy:
hyperfractionation, 81.6Gy total dose.(1.2Gy twice
daily to a total dose of 81.6Gy)
The RTOG study 83-02 [1] was a randomized Phase I/II
study of escalating doses for Hyperfractionated
radiotherapy (HF, 1.2Gy twice daily to doses of 64.8,
72, 76.8, or 81.6Gy) and Accelerated
Hyperfractionated radiotherapy (AHF, 1.6Gy twice daily
to doses of 48 or 54.4Gy) with carmustine (BCNU) for
adults with supratentorial glioblastoma multiforme
(GBM) or anaplastic astrocytoma.
The study has revealed that GBM patients who received
the higher HF doses had survival superior to the
patients in the AHF arms or lower HF doses.
In silico experiments corresponding to the different
arms of RTOG 83-02 study have been performed by
ISOG/ICCS/NTUA [17].
P
RADIOBIOLOGICAL DATA BASED ON THE
GBM GENETIC PROFILE (FROM
EXPERIMENTAL WORK)
In the results presented here, a hypothetical GBM
tumour with mutant (mt) p53 gene is considered
[2]: αp= 0.17 Gy-1, βP =0.02 Gy-2
(we also set αG0 = αP /OER, βG0 = βP /OER2, OER = 3,
and αS = 0.6 αP + 0.4 αG0 , βS = 0.6 βP + 0.4 βG0
,[3],[4] (p.99)).
The meaning of the symbols used is the following:
αp, βP : the LQ Model parameters for all
proliferative cell cycle phases except for the DNA
synthesis phase (S phase).
αS, βS : the LQ Model parameters for the S phase.
αG0, βG0 : the LQ Model parameters for the resting
G0 phase.
Typical clonogenic cell densities are 104 to 105 cells/mm3
[5].
Since most GBM tumours are poorly differentiated and
rapidly growing, we assume a clonogenic cell density of
2105 cells/mm3 in the proliferating cell region, 105
cells/mm3 in the G0 cell region and 0.2105 cells/mm3 in
the dead cell region of the tumour.
The cell cycle duration has been taken equal to 40h. This
is the average of the cell cycle durations we have found
in the literature for GBM cell lines [6],[7].
In [8] the approximate percentage of the cell cycle time
spent in each phase by a typical malignant cell is assumed
as follows: TG1 = 0.4 TC, TS = 0.39TC, TG2 = 0.19TC, TM =
0.02TC. The duration of the G0 phase is taken to be
TG0=25h [9].
The cell loss factor (CLF) is considered equal to
0.3 [10]. In [11] the authors note that cell loss is
mainly due to necrosis (CLFN) and apoptosis
(CLFA) and that gliomas have a low CLF in
general.
We assume that the total CLF (0.3) is the sum of
the CLFN (0.27) and CLFA (0.03).
We hypothesize low levels of apoptotic cells for
GBM, as we have found that this is in general the
case for gliomas [2],[11],[12]. )
CLINICAL TRIAL SPECIFICS
The delivery of irradiation takes place at 08:00
and 16:00 every day, 5 days per week (no
irradiation during weekends).
The distribution of the absorbed dose in the
tumour region is assumed to be uniform.
It should also be noted that carmustin, which
was administered to all patients enrolled in the
RTOG – 83-02 study, is assumed not to
significantly modify the relative effectiveness of
the radiation therapy schedules considered, as
the chemotherapy administration schedule was
the same for all patients.
Figure 1 presents the results of the in silico
experiments in the form of the number of
surviving tumour cells (proliferating and
dormant) as a function of time (Last time point:
8 weeks after the beginning of the radiotherapy
treatment, t=1344h).
Improved tumour control following high-dose
HF irradiation is evident in the diagram and in
agreement with the conclusions of the clinical
trial.
In fact (data not shown here), the higher the
total dose in an HF schedule, the better the
result in terms of tumour cell kill.
GBM with mutant p53
1.0E+11
1.0E+10
Number of alive tumour cells
...
1.0E+09
1.0E+08
1.0E+07
AHF- 48Gy
1.0E+06
HF- 81.6Gy
1.0E+05
1.0E+04
1.0E+03
1.0E+02
1.0E+01
1.0E+00
0
1
2
3
4
5
6
7
8
Time (weeks)
Figure 1. Number of surviving tumour cells as a function of time for a
glioblastoma tumour with mutant p53 gene. AHF-48Gy: accelerated
hyperfractionation, 48Gy total dose, HF-81.6Gy: hyperfractionation,
81.6Gy total dose.
Initial tumour region at the beginning of the radiotherapy
treatment
8 weeks later (t=1344h)
Irradiation according to
AHF 48Gy
8 weeks later (t=1344h)
Irradiation according to
HF 81.6Gy
Colour code → red: proliferating cell region, green: G0 cell region,
blue: dead cell region. See [17] for details
More specifically, the inspection of the
simulation results reveals that AHF
schedules, which employ a higher fraction
dose compared to HF schedules, seem at
first to be beneficial as they achieve the
maximum tumour cell kill at some instant.
Nevertheless, the duration of the AHF
schedules is smaller; as a result, if they
fail in eradicating “all” tumour cells,
tumour repopulation begins earlier.
Descriptions of various aspects of the
“top-down” simulation approach
developed at the In Silico Oncology Group
Institute of Communication and Computer
Systems, Natl. Tech.Univ.of Athens
[www.in-silico-oncology.iccs.ntua.gr ]and
applied on the clinical cases considered
can be found in publications [13-16] and
in particular in [17].
Need for code execution on Grid
environment
Use of Grid architectures is important in
order to execute the code for a very large
number of parameters values
combinations concurrently.
This is necessary in order to increase the
reliability of the “Oncosimulator” as many
parameter values are not known but only
their ranges can be estimated.
REFERENCES
[1] Werner-Wasik M. et al “Final report of a phase I/II trial of
hyperfractionated and accelerated hyperfractionated radiation therapy
with carmustine for adults with supratentorial malignant gliomas”,
Cancer 77, 1535-1543, 1996.
[2] D.A. Haas-Kogan, G. Yount, M. Haas, D. Levi, S.S. Kogan, L. Hu, C.
Vidair, D.F. Deen, W.C. Dewey, M.A. Israel, p53-dependent G1 arrest and
p53 independent apoptosis influence the radiobiologic response of
glioblastoma, Int. J. Radiat. Oncol. Biol. Phys. 36, 95-103, 1996.
[3] M. Kocher, H. Treuer, J. Voges, M. Hoevels, V. Sturm, R.P. Mueller,
Computer simulation of cytotoxic and vascular effects of radiosurgery in
solid and necrotic brain metastases, Radiother. Oncol. 54, 149-156,
2000.
[4] C. Perez and L. Brady, Principles and Practice of Radiation Oncology.
Philadelphia: Lippincott-Raven, 1998.
[5] B. Jones and R.G. Dale, Mathematical models of tumour and normal
tissue response, Acta Oncol. 38, 883-893, 1999.
[6] L.E. Dillehay, A model of cell killing by low-dose-rate radiation
including repair of sublethal damage, G2 block, and cell division, Rad.
Res. 124, 201-207, 1990.
REFERENCES (cont.)
[7] B. Hegedues, A. Czirok, I. Fazekas, T. Babel, E. Madarasz, T. Viscsek,
Locomotion and proliferation of glioblastoma cells in vitro: statistical evaluation
of videomicroscopic observations, J. Neurosurgery 92, 428-434, 2000.
[8] B.G. Katzung (ed), Basic and Clinical Pharamacology. International Edition.
2001.
[9] W. Duechting, W. Ulmer, R. Lehrig, T. Ginsberg, E. Dedeleit, Computer
simulation and modelling of tumour spheroid growth and their relevance for
optimization of fractionated radiotherapy, Strahlenther. Onkol. 168, 354-360,
1992.
[10] P. Huang, A. Allam, L. Perez, A. Taghian, J. Freeman, H. Suit, The effect of
combining recombinant human tumor necrosis factor-alpha with local radiation on
tumor control probability of a human glioblastoma multiforme xenograft in nude
mice, Int J Radiat Oncol Biol Phys 32(1), 93-98, 1995.
[11] M. Nakajima, S. Nakasu, S. Morikawa, T. Inubushi, Estimation of volume
doubling time and cell loss in an experimental rat glioma model in vivo, Acta
Neurochir 140, 607-613, 1998.
[12] S. Tribius, A. Pidel, D. Casper, ATM protein expression correlates with
radioresistance in primary glioblastoma cells in culture, Int. J. Radiat. Oncol.
Biol. Phys. 50, 511-523, 2001.
REFERENCES (cont.)
[13] G.S.Stamatakos, D.D.Dionysiou, E.I.Zacharaki, N.A.Mouravliansky, K.Nikita,
N.Uzunoglu, “In silico radiation oncology: combining novel simulation algorithms
with current visualization techniques”, Proceedings of the IEEE, vol. 90, No11,
Nov. 2002. 1764-1777
[14] D. D. Dionysiou, G. S. Stamatakos, N.K. Uzunoglu, K. S. Nikita, A. Marioli, “A
four-dimensional simulation model of tumour response to radiotherapy in vivo:
parametric validation considering radiosensitivity, genetic profile and
fractionation,” Journal of Theoretical Biology 230 (2004) 1–20
[15] V. P Antipas, G. S Stamatakos, N. K Uzunoglu, D. D Dionysiou, R. G Dale, ” A
spatio-temporal simulation model of the response of solid tumours to radiotherapy
in vivo: parametric validation concerning oxygen enhancement ratio and cell cycle
duration,” Phys. Med. Biol. 49 (2004) 1485–1504
[16]G.S.Stamatakos GS, V.P. Antipas VP, N.K. Uzunoglu, “Simulating
chemotherapeutic schemes in the individualized treatment context: The paradigm
of glioblastoma multiforme treated by temozolomide in vivo.” Comput Biol Med.
2005 Oct 2; [Epub ahead of print, Pubmed Link:
http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=pubmed&dopt
=Abstract&list_uids=16207487&query_hl=14]
[17] G. S. Stamatakos, V.P. Antipas, N. K. Uzunoglu, R. G. Dale, “A four
dimensional computer simulation model of the in vivo response to radiotherapy of
glioblastoma multiforme: studies on the effect of clonogenic cell density.” British
Journal of Radiology, 2006, vol. 79, 389-400
Thank you