=Paper=
{{Paper
|id=Vol-2086/AICS2017_paper_13
|storemode=property
|title=A Computational Lymph Tissue Model for Long Term HIV Infection Progression and Immune Fitness
|pdfUrl=https://ceur-ws.org/Vol-2086/AICS2017_paper_13.pdf
|volume=Vol-2086
|authors=Andreas Hillmann,Martin Crane,Heather Ruskin
|dblpUrl=https://dblp.org/rec/conf/aics/HillmannCR17
}}
==A Computational Lymph Tissue Model for Long Term HIV Infection Progression and Immune Fitness==
https://ceur-ws.org/Vol-2086/AICS2017_paper_13.pdf
A computational lymph tissue model for long term HIV
infection progression and immune fitness
Andreas Hillmann1, Martin Crane1 and Heather J. Ruskin1
1 Advanced Research Computing Centre for Complex Systems Modelling (ARC-SYM),
School of Computing, Dublin City University, Dublin, Ireland
andreas.hillmann2@mail.dcu.ie
Abstract. Given the complexity of biological systems, modelling the immune
response to the spread of infectious disease is non-trivial and has been the subject
of considerable computational efforts. HIV infection, if untreated, leads to a slow
depletion of the immune system over many years, resulting in terminal AIDS
disease and death. Despite decades of research, the biological mechanisms un-
derpinning the immune system response and implicit in its impairment remain a
subject of discussion. Recent biological findings indicate the importance of func-
tional lymphatic tissue in maintaining homoeostasis of the immune system,
which is vital for stability. HIV infection-related immune activation leads to
chronic inflammation of this tissue and formation of non-functional scar tissue
with detrimental effects on homeostasis. We propose a computational model
based on the Cellular Automata formalism to quantify these effects in a simulated
fraction of a lymphatic tissue. We also include effects of antiretroviral treatment
and highlight implications for system representation through larger-scale simu-
lations, as well as optimizations of treatment schedules.
Keywords: HIV, computational model, Cellular Automata, antiretroviral treat-
ment
1 Introduction
1.1 HIV infection
Acquired Human Immunodeficiency Syndrome (AIDS) remains a global health con-
cern even three decades or more after the discovery of the Human Immunodeficiency
Virus as its root cause, [1]. The virus preferentially infects white blood cells of a certain
type (CD4+), which are a crucial part of the immune response, leading to failure of the
immune system approximately 10 years after initial infection. Potent drugs, termed An-
tiretroviral Therapy (ART), have been developed, which aim at the suppression of viral
activities. However, these medications do not provide a cure for the disease, due to the
persistence of impermeable viral reservoirs in the body; instead they have to be admin-
istered life-long to maintain sufficient viral suppression, [2]. Daily administration of
multiple drugs can become a problem and interruptions are common, especially in re-
source-limited settings, where drug availability is not guaranteed or social factors in-
terfere with adherence, [3].
Moreover, Structured Treatment Interruptions (STIs) have been suggested to enable
recovery from drug toxicities, to promote resilience of the immune system and to reduce
drug costs, [4]. Though initial clinical studies on STIs provided encouraging evidence
regarding safety of interrupted treatment, [5, 6], a subsequent large-scale study
(SMART) indicated potential harmful effects, [7]. Despite these effects have not been
characterized in detail and may have been related to study design, [8], the results led to
a general loss of interest in assessing treatment interruptions in a clinical setting. Ques-
tions regarding safe margins of interruption regimen and the nature of detrimental ef-
fects thus remain open.
1.2 Lymphatic tissue involvement
It is well-known that viral replication occurs predominantly in lymphatic tissue, where
most of the susceptible cells are located, [9]. The lymphatic system is a widespread
network of nodes and ducts which is crucial for immune response but also for the spread
of infected cells throughout the body.
Recent advances in in-vivo microscopy have enabled determination of the detailed
structure of a certain kind of lymphatic tissue, the Fibroblastic Reticular Cell Network
(FRCn). Its mesh-like structure has been reported to mimic a small world topology with
lattice-like properties, [10]. This view is complemented by histological imaging show-
ing that the Fibroblastic Reticular Cells (FRCs) are fixed to a scaffolding of collagen
fibers. During inflammation, this formation of collagen fibers is greatly increased, [11],
leading to collagen deposits (see Fig. 1) which block off FRCs from the chemical sig-
nals (cytokines) which maintain their viability, leading to depletion of these cells and
formation of non-functional collagen tissue. Antiretroviral treatment has been shown
to stop the formation of the scar tissue but not to enable its reversal, [12]. The formation
of FRCn has not yet been observed in vivo due to the long timescales involved (weeks
to years), however.
Fig. 1. Distinct stages of collagenation of lymphatic tissue for uninfected, early, and late stage
HIV infection (from left). Green staining denotes functional tissue, red non-functional collagen.
Image source: [12].
1.3 Modelling approaches
Various models have been proposed in the HIV context to help understanding the
spread of the virus between, [13], and within hosts, [14]. The majority of these models
have found to be based on systems of Ordinary Differential Equations (ODEs), [15],
representing a variant of the SIR model originally used to model spread of communi-
cable diseases throughout populations. In context of within-host models, each DE de-
notes the change of a quantitative property related to HIV like viral load, healthy and
infected cell concentrations over time. Starting from basic forms with two or three equa-
tions, models have been extended to cover additional features, like effects of antiretro-
viral treatment or immune response (resulting in systems of 20 or more equations).
Simpler forms allow further mathematical analysis, [16], whereas more complex forms
may be used for numerical simulation studies, [17]. For a more detailed review on ODE
based HIV-modelling see [14] and references therein.
In general, ODE-based approaches assume virus und susceptible cells being in an
environment satisfying the conditions of well-mixed. That assumption fits loosely best
to the bloodstream, which initially has been considered a main source of viral replica-
tion, [14]. However, following the increasing availability of data on tissue involvement,
the suitability of the well-mixed assumption has been questioned recently, [18].
Treatment interruptions and optimizations have also been subject to modelling stud-
ies using ODEs, [15]. A popular approach, besides pure simulation studies, [19], con-
siders an optimal control problem, [20–22], to obtain treatment regimen which keep
biological properties, like viral load or healthy cell count within predefined bounds
which are considered to be non-critical. This view of the biological system as a kind of
thermostat was especially useful with only limited data on anatomical compartments
other than the bloodstream available. However, as clinical studies have indicated the
complexity of HIV infection progression, consideration of alternative modelling ap-
proaches might be worthwhile.
We therefore propose a computational model, which considers both spatial structure
of lymphatic tissue and the effects of inflammation, to examine long-term development
and to allow predictions to be made on effects of treatment interruptions, where clinical
data is not available for reasons mentioned in section 1. Our model presents an exten-
sion of the approach reported in [15] with improved biological realism obtained from
drawing upon recent studies.
2 Methods - Numerical simulations
2.1 Surrounding conditions
For our model we assume two effects to be responsible for the depletion of CD4+ cells,
as reported by clinical studies, [11].
• HIV-induced direct killing of susceptible CD4+ cells
• Impairment of survival through depletion of FRCs
Both factors influence the concentration of CD4+ cells in the body, with the time-scale
of the former relatively short (days to weeks) while the latter is effective at longer time-
scales (months to years). However due to regenerative capabilities of CD4+ cells, the
first effect is reversible under antiretroviral treatment, while only minimal regeneration
has been observed, [12], for the second.
We apply a bottom-up modelling approach, by using the Stochastic Cellular Autom-
ata (SCA) formalism, i.e. a stochastic extension of Cellular Automata (CA), [23]. The
CA formalism has successfully been applied for modelling environments involving
fixed matrices, including HIV, [24]. It utilizes a regular grid or lattice structure, where
each site represents a single cell and its neighborhood the cell-to-cell contacts. We fur-
ther assume that each cell has a connection to its 4 immediate neighbors as suggested
from analysis of in-vivo imaging data, [10]. In terms of Cellular Automata computation,
this corresponds to a von Neumann neighborhood.
An average lymph node measures about 1 cm in diameter and is of approximately
spherical shape. From murine experiments, it is known that in FRCn, mean separation
of the centers of mass of single cells is ~ 23 µm, [10]. A human lymph node, would
therefore consist of around 40 Million FRCs. For our simulations, however, we con-
sider a smaller fraction, namely a 2-dimensional slice through a lymph node, consisting
approximately of 250,000 single cells. This slice represents less than 1% of the total
volume, however this simplified view is acceptable in this case since we do not aim at
reproducing an anatomically correct representation of the organ (see e.g. [25]).
From clinical data, we know that collagenation mainly starts from connections in the
blood circulation (High Endothelial Venules). These connections appear to be scattered
randomly throughout the tissue, [12]. We assume that collagenation increase occurs
predominantly through elongation of existing collagen fibers as suggested by those data
available to date on this aspect.
2.2 Model rules
Before simulated infection occurs, the lattice is populated with both healthy FRCs and
co-located CD4+ cells. The CD4+ cells are assumed to maintain equilibrium (homeo-
stasis) and their concentrations in lymphatic tissue to be stable, [26]. However, HIV
infection has a two-fold effect in our model: destroying CD4+ cells through viral infec-
tion and mechanisms of cellular apoptosis and pyroptosis, and through upregulation of
the immune system, [26]. The upregulated immune response in turn causes increased
collagenation of lymphatic tissue.
The fraction of CD4+ cells which is co-located with collagenated tissue, tends to
exhibit poorer survival than usual (in healthy tissue), in addition to the depletion due to
HIV infection. Treatment, however reduces the latter, as well as inflammation stimula-
tion, since viral load is reduced. CD4+ cells are therefore regenerated and the colla-
genation processes in lymphatic tissue are slowed down. Upon reinstatement of upreg-
ulation of the immune system due to a treatment interruption, however, homeostasis
disruption occurs again.
In terms of model rules, no direct death and/or inflammation takes place other than
by viral infection. This context gives rise to the state transition model shown in Fig. 1,
which contains the four possible combinations of CD4+ cells and collagen status as
well as permitted changes.
Fig. 2. Block diagram of state changes of the SCA with probabilities of transitions, where appli-
cable (for details see Table 1.). H: healthy CD4+ cell; HC: healthy CD4+ cell co-located with
collagen; D: dead CD4+ cell (empty); C: empty site co-located with collagen.
As mentioned in the previous section, collagenation originates from the High Endothe-
lial Venules, where CD4+ cells enter the lymph nodes from the blood circulatory sys-
tem. The assumption, that the blood circulation acts as a means to spread infection
throughout the body, allows us to calibrate the model using available clinical data on
CD4+ cell long-term degradation under HIV infection, obtained from blood samples,
[27]. Table 1 lists the model parameters together with values obtained by calibration.
Table 1. Model parameter value selection
Parameter Description Default value
PT Treatment effectiveness 0.9
PINF Infection induced inflammation, drives collagenation 0.1
PREG CD4+ cell regeneration 0.3
PCYT CD4+ cell death due to signaling (cytokines) 0.9
PAP CD4+ cell death through infection (apoptosis) 0.1
N Collagenated neighbors 1-4
2.3 Simulation setup
Using the model parametrization in this table, we aim to test the effect of different
patterns of unstructured interrupted treatment in the long term on collagenation (and
thus immune system degradation). To simulate interrupted treatment, we assume ‘av-
erage’ adherence (the probability that the simulated patient takes the medication each
week). As the treatment initiation timing has also been deemed to be important in clin-
ical studies, [28], we allow for this in the simulations as well. Table 2. summarizes the
parameter values chosen for the simulation runs. After the treatment period (corre-
sponding to ~ 5 years), the increase in collagenation compared to that evident at the
start of treatment is determined.
Table 2. Simulation parameters
Description Value range
Time step 1 week
Treatment initiation 0, 50, 100, 150, 200 time steps
Treatment adherence 0%, 20%, 40%, 60%, 80% 100%
Treatment period duration 250 time steps
Total simulation length 500 time steps (~ 10 years)
In a second experiment, we aim to investigate the effects of structured treatment inter-
ruptions. Thus, we adopt the approach of CD4+ guidance, which has been pursued in
recent clinical trials, [4].This approach uses a threshold value of CD4+ cell count in a
patient as a surrogate marker to decide on interruption or (re-)initiation of antiretroviral
treatment. For our simulations, we apply the limits used in two major clinical trials, [7,
29] and again observe the outcomes after 500 time steps.
3 Results and Discussion
3.1 Unstructured interruptions
An example model run is depicted in Fig. 3, exhibiting unstructured treatment inter-
ruptions due to incomplete adherence. In general, the model runs demonstrate good
correspondence to clinical results, [6]. For an infection simulated to occur at time step
0, the count of CD4+ cells drops rapidly, (characteristic of the acute phase). The cell
count continues to drop at a slower rate until initiation of simulated treatment, which
leads to a significant recovery in number of CD4+ cells. Incomplete adherence to the
treatment regime causes temporary drops in cell numbers (signified by the irregular saw
tooth pattern), as well as an overall decline in the long term. This decline accelerates
once the treatment period concludes. Collagenation, however, exhibits a constant in-
crease, albeit with less-pronounced slope during treatment.
In Fig. 4 we show the loss of CD4+ cells and the increase in collagenation during
treatment periods. The results illustrate the influence of treatment initiation timing and
of adherence to the regimen. We observe that the latter has a fundamental impact on
long-term progression of collagenation and, consequently, on immune system fitness.
Collagen CD4+ cells
100%
90%
80%
tissue area covered
70%
60%
50%
40%
30%
20%
10%
0%
0 100 200 300 400 500
time step [weeks]
Fig. 3. Time course of collagenation (blue line) and population of CD4+ cells (red line) during a
period of interrupted treatment (100 to 350 time steps; 0.8 adherence)
0.0 0.2 0.4 0.6 0.8 1.0
80%
70%
collogenation increase
60%
50%
40%
30%
20%
10%
0%
0 50 100 150 200
time step of treatment initiation [weeks]
Fig. 4. Effect of different adherence levels (lines) and treatment initiation on collagenation in-
crease at end of treatment period.
Treatment initiation timing has a less-pronounced effect. Results show that, if treatment
is introduced during the early phases of infection (within the first 2 years), initiation
time has greater impact than for later stages. However, with high adherence (80% and
above), collagenation can be kept within acceptable bounds even for late treatment in-
itiation. As an acceptable outcome, we take a total collagenated area of less than 75%
at the end of the simulated 10 years. This level has been determined as a threshold from
clinical data, [12], above which AIDS disease is imminent.
3.2 Structured interruptions
Example results of our second experiment concerning structured treatment interrup-
tions are shown in Fig. 5. Using the same thresholds for CD4+ guidance as two large-
scale clinical studies (SMART, LOTTI, see Table 3.), simulated data is in good corre-
spondence with clinical data. The two sets of clinical data indicate that counts of CD4+
cells are constantly decreasing during the course of these studies, reflected by the sim-
ulation data. Since the duration of the clinical trials was limited due to practical reasons,
we extended simulation times to allow for an estimation of long-term effects of the
interruption thresholds used in the studies.
LOTTI SMART
1200
CD4+ count [cells/µl]
1000
800
600
400
200
0 10 20 30 40 50
study duration [months]
Fig. 5. Comparison of clinical data from two CD4+ guided STI studies (LOTTI, SMART; dots)
with model output (lines) from corresponding simulation experiments.
On this behalf, absolute values of CD4+ count limits from clinical literature were nor-
malized for use with the model by assuming an initial cell count of 1000 cells/µl.and
results were again obtained after 500 time steps (~ 10 years). The results (see Table 3.)
show a relatively small difference both in terms of collagenation and CD4+ counts.
These results are remarkable in a way since the SMART study strongly indicated ad-
verse effects of interrupted treatment, whereas the LOTTI study did not find significant
differences in this regard compared to continuous treatment. The different outcomes of
our simulations in terms of collagenation and CD4+ count may therefore account for
some threshold between non-critical progression and adverse effects.
Table 3. Experimental results for CD4+ guided structured treatment interruptions
CD4+ count limits [cells/µl] Simulation results (~ 10 years)
Clinical study Interruption Resumption Collagenation CD4+ count
SMART [7] 350 250 97% 260
LOTTI [29] 700 350 81% 380
4 Conclusion
We have proposed a simple Stochastic Cellular Automata model, which describes the
effects of HIV infection and antiretroviral therapy on both CD4+ cell counts and colla-
gen deposition in lymphatic tissue. Additionally, the model can predict possible out-
comes of incomplete antiretroviral treatment, both unstructured and structured (difficult
to obtain from controlled clinical studies). Simulated adherence patterns indicated the
importance for long-term immunological fitness, of overall adherence to medication,
and of treatment initiation timing during the early phases of infection, with regard to
improved management. These experiments on antiretroviral treatment timing and ad-
herence explore the implications for outcomes of Structured Treatment Interruptions,
formerly discounted due to lack of long term predictions, despite some evident patient
benefits.
Further steps might include an extension of our approach to evaluate strategies for
an improved monitoring of antiretroviral therapy by helping to identify patterns, like
certain concentration thresholds, which may indicate a future loss of viral control given
poor adherence by employing AI techniques. Applicability in real-life settings may fur-
therly be improved by incorporation of realistic sampling intervals to obtain biological
properties from blood or tissue with least patient stress.
Regarding structured interruptions, numerical simulation in conjunction with appro-
priate optimization techniques could be utilized for identifying optimal treatment regi-
men with improved biological realism compared to ODE-based approaches. A conceiv-
able way would be the use of evolutionary algorithms to optimize treatment on/off pat-
terns.
In addition, the proposed model offers further development potential, not least as it
represents to date only a small part of human lymphatic tissue. A realistic simulation
size even for one lymph node requires around 100 times more lattice sites than the
current model. This problem can be alleviated using state-of-the-art parallelization ap-
proaches, [30], to simulate different regions and this forms part of ongoing investiga-
tion.
Unfortunately, clinical data for validation of model performance are sparse at present
due to ethical considerations as well as measurement limitations, but recent advances
in biological imaging and in-vivo microscopy may well provide additional insights and
enable refinement of key parameters.
In summary, computational modelling of lymphatic tissue features involving cell-
to-cell connections offers considerable potential for further investigation and optimiza-
tion of long-term HIV treatment. Promising foundations using bottom-up modelling
have already been laid for investigating infectious diseases. We think it worthwhile to
pursue further, given growing availability of appropriate experimental and clinical data.
References
1. Barré-Sinoussi, F., Chermann, J.C., Rey, F., Nugeyre, M.T., Chamaret, S., Gruest, J.,
Dauguet, C., Axler-Blin, C., Vézinet-Brun, F., Rouzioux, C., Rozenbaum, W.,
Montagnier, L.: Isolation of a T-lymphotropic retrovirus from a patient at risk for
acquired immune deficiency syndrome (AIDS). Science. 220, 868–871 (1983).
2. WHO: Consolidated guidelines on the use of antiretroviral drugs for treating and
preventing HIV infection,
http://www.who.int/hiv/pub/guidelines/arv2013/download/en/.
3. Mann, M., Lurie, M.N., Kimaiyo, S., Kantor, R.: Effects of political conflict-induced
treatment interruptions on HIV drug resistance. AIDS Rev. 15, 15–24 (2013).
4. Benson, C.A.: Structured treatment interruptions--new findings. Top. HIV Med. 14,
107–11 (2006).
5. Rosenberg, E.S., Altfeld, M., Poon, S.H., Phillips, M.N., Wilkes, B.M., Eldridge, R.L.,
Robbins, G.K., D’Aquila, R.T., Goulder, P.J., Walker, B.D.: Immune control of HIV-1
after early treatment of acute infection. Nature. 407, 523–526 (2000).
6. Dybul, M., Chun, T.W., Yoder, C., Hidalgo, B., Belson, M., Hertogs, K., Larder, B.,
Dewar, R.L., Fox, C.H., Hallahan, C.W., Justement, J.S., Migueles, S.A., Metcalf, J.A.,
Davey, R.T., Daucher, M., Pandya, P., Baseler, M., Ward, D.J., Fauci, A.S.: Short-cycle
structured intermittent treatment of chronic HIV infection with highly active
antiretroviral therapy: effects on virologic, immunologic, and toxicity parameters. Proc.
Natl. Acad. Sci. U. S. A. 98, 15161–6 (2001).
7. Siegel, L., El-Sadr, W.: New perspectives in HIV treatment interruption: The SMART
study. PRN Noteb. 11, 8–9 (2006).
8. Hirschel, B., Flanigan, T.: Is it smart to continue to study treatment interruptions? AIDS.
23, 757–759 (2009).
9. Pantaleo, G., Graziosi, C., Demarest, J.F., Butini, L., Montroni, M., Fox, C.H.,
Orenstein, J.M., Kotler, D.P., Fauci, A.S.: HIV infection is active and progressive in
lymphoid tissue during the clinically latent stage of disease. Nature. 362, 355–358
(1993).
10. Novkovic, M., Onder, L., Cupovic, J., Abe, J., Bomze, D., Cremasco, V., Scandella, E.,
Stein, J. V., Bocharov, G., Turley, S.J., Ludewig, B.: Topological Small-World
Organization of the Fibroblastic Reticular Cell Network Determines Lymph Node
Functionality. PLOS Biol. 14, e1002515 (2016).
11. Schacker, T.W., Reilly, C., Beilman, G.J., Taylor, J., Skarda, D., Krason, D., Larson,
M., Haase, A.T.: Amount of lymphatic tissue fibrosis in HIV infection predicts
magnitude of HAART-associated change in peripheral CD4 cell count. AIDS. 19, 2169–
2171 (2005).
12. Zeng, M., Southern, P.J., Reilly, C.S., Beilman, G.J., Chipman, J.G., Schacker, T.W.,
Haase, A.T.: Lymphoid Tissue Damage in HIV-1 Infection Depletes Naïve T Cells and
Limits T Cell Reconstitution after Antiretroviral Therapy. PLoS Pathog. 8, e1002437
(2012).
13. May, R.M., Anderson, R.M.: Transmission dynamics of HIV infection. Nature. 326,
137–142 (1987).
14. Perelson, A.S., Nelson, P.W.: Mathematical Analysis of HIV-1 Dynamics in Vivo.
SIAM Rev. 41, 3–44 (1999).
15. Hillmann, A., Crane, M., Ruskin, H.J.: HIV models for treatment interruption:
Adaptation and comparison. Phys. A Stat. Mech. its Appl. 483, 44–56 (2017).
16. Pawelek, K. a., Liu, S., Pahlevani, F., Rong, L.: A model of HIV-1 infection with two
time delays: Mathematical analysis and comparison with patient data. Math. Biosci. 235,
98–109 (2012).
17. Sun, Q., Min, L.: Dynamics Analysis and Simulation of a Modified HIV Infection
Model with a Saturated Infection Rate. Comput. Math. Methods Med. 2014, 1–14
(2014).
18. Cohen, J.: Tissue Says Blood Is Misleading, Confusing HIV Cure Efforts. Science (80-
. ). 334, 1614–1614 (2011).
19. Ferreira, J., Hernandez-Vargas, E. a., Middleton, R.H.: Computer simulation of
structured treatment interruption for HIV infection. Comput. Methods Programs
Biomed. 104, 50–61 (2011).
20. Adams, B.M., Banks, H.T., Kwon, H.-D., Tran, H.T.: Dynamic multidrug therapies for
hiv: optimal and sti control approaches. Math. Biosci. Eng. 1, 223–241 (2004).
21. Krakovska, O., Wahl, L.M.: Optimal drug treatment regimens for HIV depend on
adherence. J. Theor. Biol. 246, 499–509 (2007).
22. Rivadeneira, P.S., Moog, C.H., Stan, G.-B., Brunet, C., Raffi, F., Ferré, V., Costanza,
V., Mhawej, M.J., Biafore, F., Ouattara, D. a., Ernst, D., Fonteneau, R., Xia, X.:
Mathematical Modeling of HIV Dynamics After Antiretroviral Therapy Initiation: A
Review. Biores. Open Access. 3, 233–241 (2014).
23. von Neumann, J.: Theory of self-reproducing automata. Inf. Storage Retr. 5, 151 (1969).
24. Wodarz, D., Levy, D.N.: Effect of different modes of viral spread on the dynamics of
multiply infected cells in human immunodeficiency virus infection. J. R. Soc. Interface.
8, 289–300 (2011).
25. Kislitsyn, A., Savinkov, R., Novkovic, M., Onder, L., Bocharov, G.: Computational
Approach to 3D Modeling of the Lymph Node Geometry. Computation. 3, 222–234
(2015).
26. Fletcher, A.L., Acton, S.E., Knoblich, K.: Lymph node fibroblastic reticular cells in
health and disease. Nat. Rev. Immunol. 15, 350–361 (2015).
27. Fauci, A.S., Pantaleo, G., Stanley, S., Weissman, D.: Immunopathogenic mechanisms
of HIV infection. Ann. Intern. Med. 124, 654–663 (1996).
28. Lundgren, J.D., Babiker, A., Gordin, F., Emery, S., Grund, B., Sharma, S.,
Avihingsanon, A., Cooper, D.A., Fätkenheuer, G., Llibre, J.M., Molina, J.-M., Munderi,
P., Schechter, M., Wood, R., Klingman, K.L., Collins, S., Lane, H.C., Phillips, A.N.,
Neaton, J.D.: Initiation of Antiretroviral Therapy in Early Asymptomatic HIV Infection.
N. Engl. J. Med. 373, 795–807 (2015).
29. Maggiolo, F., Airoldi, M., Callegaro, A., Martinelli, C., Dolara, A., Bini, T., Gregis, G.,
Quinzan, G., Ripamonti, D., Ravasio, V., Suter, F.: CD4 cell-guided scheduled
treatment interruptions in HIV-infected patients with sustained immunologic response
to HAART. AIDS. 23, 799–807 (2009).
30. Perrin, D., Bezbradica, M., Crane, M., Ruskin, H.J., Duhamel, C.: High-Performance
Computing for Data Analytics. 2012 IEEE/ACM 16th Int. Symp. Distrib. Simul. Real
Time Appl. 234–242 (2012).