One in five women over the age of 50 will experience a fragility fracture in their remaining lifetime (International Osteoporosis Foundation). The anti-resorptive drug denosumab significantly increases bone mass and reduces the risk of fracture. It is the second most widely used prescription medicine for the treatment of post-menopausal osteoporosis. In vivo experiments in the field of bone biome- chanics are time-consuming and expensive so in silico models provide a powerful tool to test hypotheses concerning bone remodeling and compare treatment and placebo scenarios for the same patient. In the first part of this work, we present an in-silico model of the development of osteoporotic iliac crest bone with and without denosumab treatment. Estrogen signaling directly to osteoclasts, osteoblasts and osteocytes and its effects on the RANKL/OPG system allowed postmenopausal osteoporosis to be in- cluded within the platform. Denosumab was modelled as an additional anti-body for RANKL. The model was validated against dynamic and morphometric parameters observed in-vivo during clinical studies of denosumab. Seven biopsies were selected covering the whole range of ages and bone volume fractions in the clinical studies. The denosumab-treated simulations after 2.5, 5 and 10 years showed non-significant variations from clinical results in bone volume fraction (p=0.042, 0.0046 and 0.043 respectively). Dynamic parameters such as osteoclast numbers, mineralizing and eroded surfaces qualitatively followed the trends observed in clinical trials with a strong decrease in osteoclast numbers and eroded surface soon after the first denosumab injection and a smaller but also significant decrease in mineralising surface. The validity of the model was further confirmed by examination of the microstructural changes and the cellular behavior at three different scales: single trabeculae (dimensions of section: 525 μm by 525 μm by 525 μm), groups of trabeculae (dimensions: 1050 μm by 1050 μm by 1050 μm) and complete biopsies (dimensions: 2961 μm by 2961 μm by 2772 μm). The model was run for a limited group of biopsies for a 10-year period, in which at regular time periods new treatment groups were branched from the control simulations. The resulting data will be used to generate a decision tree regarding the timing of interventions. In the second part of this work, we present a parallel processing algorithm that significantly reduced the minimal computational time achievable for a given biopsy size. We present the dependence of the compu- tational time on the number of nodes, cores and processors used and propose a method to determine the optimal combination of nodes, cores and processors for each problem size. This algorithm facilitates the high resolution modeling of large bone biopsies over extended time periods and allows for the inclusion of up to 30 different chemical factors involved in as many reactions.
Una donna su cinque di età superiore ai 50 anni subirà una frattura da fragilità nella loro vita residua (International Osteoporosis Foundation [1]). Il farmaco anti-riassorbimento denosumab aumenta in modo significativo la massa ossea e riduce il rischio di frattura [2]. È il secondo farmaco di prescrizione più utilizzato per il trattamento dell’osteoporosi post-menopausale [3]. Esperimenti in vivo sull’osso sono lunghi e costosi, quindi i modelli in silico forniscono un potente strumento per testare ipotesi riguardanti il rimodellamento osseo e confrontare gli scenari di trattamento e placebo per lo stesso paziente. [4]. Nella prima parte di questo lavoro, presentiamo un modello in-silico dello sviluppo dell’osso osteoporotico della cresta iliaca con e senza trattamento con denosumab. L’osteoporosi postmenopausale è stata model- lata come una mancanza di estrogeni che agisce direttamente su osteoclasti, osteoblasti e osteociti e in- fluenza il sistema RANKL-OPG. Denosumab è stato modellato come un ulteriore anti-corpo per RANKL. Il modello è stato validato rispetto ai parametri dinamici e morfometrici osservati in vivo durante studi clinici su denosumab [2, 5, 6]. Sette biopsie sono state selezionate per coprire l’intera gamma di età e le frazioni del volume osseo negli studi clinici. Le simulazioni trattate con denosumab dopo 2,5, 5 e 10 anni hanno mostrato variazioni non significative dai risultati clinici nella frazione del volume osseo (p = 0,042, 0,0046 e 0,043 rispettivamente). Parametri dinamici come il numero di osteoclasti, le superfici mineralizzanti e erose hanno seguito qualitativamente le tendenze osservate negli studi clinici con una forte diminuzione del numero di osteoclasti e superficie erosa subito dopo la prima iniezione di denosumab e una minore ma significativa diminuzione della superficie mineralizzante. La validità del modello è stata ulteriormente confermata dall’esame dei cambiamenti microstrutturali e del comportamento cellulare a tre diverse scale: singole trabecole (dimensioni: 525 μm by 525 μm by 525 μm), gruppi di trabecole (dimensioni: 1050 μm by 1050 μm by 1050 μm) e biopsie complete (dimensioni: 2961 μm by 2961 μm by 2772 μm). Il modello è stato eseguito per un gruppo limitato di biopsie per un periodo di 10 anni, durante il quale in periodi di tempo regolari i nuovi gruppi di trattamento sono stati ramificati dalle simulazioni di controllo. I dati risultanti verranno utilizzati per generare un albero decisionale riguardante i tempi degli interventi. Nella seconda parte di questo lavoro, presentiamo un algoritmo di elaborazione parallela che riduce signi- ficativamente il tempo di calcolo minimo ottenibile per una data dimensione della biopsia. Presentiamo la dipendenza del tempo di calcolo sul numero di nodi, core e processori utilizzati e proponiamo un metodo per determinare la combinazione ottimale di nodi, core e processori per ciascuna dimensione del problema. Questo algoritmo facilita la modellazione ad alta risoluzione di grandi biopsie ossee su lunghi periodi di tempo e consente l’inclusione di fino a 30 diversi fattori chimici coinvolti in altrettante reazioni.
Cell-scale simulations of bone tissue : micro-multi-physics modelling of osteoporosis and its treatments
LEDOUX, CHARLES
2019/2020
Abstract
One in five women over the age of 50 will experience a fragility fracture in their remaining lifetime (International Osteoporosis Foundation). The anti-resorptive drug denosumab significantly increases bone mass and reduces the risk of fracture. It is the second most widely used prescription medicine for the treatment of post-menopausal osteoporosis. In vivo experiments in the field of bone biome- chanics are time-consuming and expensive so in silico models provide a powerful tool to test hypotheses concerning bone remodeling and compare treatment and placebo scenarios for the same patient. In the first part of this work, we present an in-silico model of the development of osteoporotic iliac crest bone with and without denosumab treatment. Estrogen signaling directly to osteoclasts, osteoblasts and osteocytes and its effects on the RANKL/OPG system allowed postmenopausal osteoporosis to be in- cluded within the platform. Denosumab was modelled as an additional anti-body for RANKL. The model was validated against dynamic and morphometric parameters observed in-vivo during clinical studies of denosumab. Seven biopsies were selected covering the whole range of ages and bone volume fractions in the clinical studies. The denosumab-treated simulations after 2.5, 5 and 10 years showed non-significant variations from clinical results in bone volume fraction (p=0.042, 0.0046 and 0.043 respectively). Dynamic parameters such as osteoclast numbers, mineralizing and eroded surfaces qualitatively followed the trends observed in clinical trials with a strong decrease in osteoclast numbers and eroded surface soon after the first denosumab injection and a smaller but also significant decrease in mineralising surface. The validity of the model was further confirmed by examination of the microstructural changes and the cellular behavior at three different scales: single trabeculae (dimensions of section: 525 μm by 525 μm by 525 μm), groups of trabeculae (dimensions: 1050 μm by 1050 μm by 1050 μm) and complete biopsies (dimensions: 2961 μm by 2961 μm by 2772 μm). The model was run for a limited group of biopsies for a 10-year period, in which at regular time periods new treatment groups were branched from the control simulations. The resulting data will be used to generate a decision tree regarding the timing of interventions. In the second part of this work, we present a parallel processing algorithm that significantly reduced the minimal computational time achievable for a given biopsy size. We present the dependence of the compu- tational time on the number of nodes, cores and processors used and propose a method to determine the optimal combination of nodes, cores and processors for each problem size. This algorithm facilitates the high resolution modeling of large bone biopsies over extended time periods and allows for the inclusion of up to 30 different chemical factors involved in as many reactions.File | Dimensione | Formato | |
---|---|---|---|
Thesis_Ledoux_Cellesi (1).pdf
non accessibile
Descrizione: Thesis text
Dimensione
4.74 MB
Formato
Adobe PDF
|
4.74 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/153024