publications

Background and purpose: Valvular Heart Disease (VHD) is a known complication of childhood cancer after radiotherapy treatment. However, the dose-volume-effect relationships have not been fully explored. Materials and methods: We obtained individual heart Dose Volume Histograms (DVH) for survivors of the French Childhood Cancer Survivors Study (FCCSS) who had received radiotherapy. We calculated the Mean Dose to the Heart (MHD) in Gy, as well as the heart DVH parameters (Vd Gy, which represents the percentage of heart volume receiving at least d Gy), fixing the thresholds to 0.1 Gy, 5 Gy, 20 Gy, and 40 Gy. We analyzed them furtherly in the subpopulation of the cohort that was treated with a dose lower than 5 Gy (V0.1Gy|V5Gy=0%), 20 Gy (V5Gy|V20Gy=0%), and 40 Gy (V20Gy|V40Gy=0%), respectively. We inves- tigated their role in the occurrence of a VHD in this population-based observational cohort study using the Cox proportional hazard model, adjusting for age at cancer diagnosis and chemotherapy exposure. Results: Median follow-up was 30.6 years. Eighty-one patients out of the 7462 (1 %) with complete data experienced a severe VHD (grade ! 3). The risk of VHD increased along with the MHD, and it was asso- ciated with high doses to the heart (V40Gy < 50 %, hazard ratio (HR) = 7.96, 95 % CI: 4.26–14.88 and V20Gy|V40Gy=0% >50 %, HR = 5.03, 95 % CI: [2.35–10.76]). Doses 5–20 Gy to more than 50 % (V5Gy|V20Gy=0% >50 %) of the heart induced a marginally non-significant estimated risk. We also observed a remarkable risk increase with attained age. Conclusions: Our results provide new insight into the VHD risk that may impact current treatments and long-term follow-up of childhood cancer survivors.

Periodic recurrence is a prominent behavioural of many biological phenomena, including cell cycle and circadian rhythms. Although deterministic models are commonly used to represent the dynamics of periodic phenomena, it is known that they are little appropriate in the case of systems in which stochastic noise induced by small population numbers is actually responsible for periodicity. Within the stochastic modelling settings automata-based model checking approaches have proven an effective means for the analysis of oscillatory dynamics, the main idea being that of coupling a period detector automaton with a continuous-time Markov chain model of an alleged oscillator. In this paper we address a complementary aspect, i.e. that of assessing the dependency of oscillation related measure (period and amplitude) against the parameters of a stochastic oscillator. To this aim we introduce a framework which, by combining an Approximate Bayesian Computation scheme with a hybrid automata capable of quantifying how distant an instance of a stochastic oscillator is from matching a desired (average) period, leads us to identify regions of the parameter space in which oscillation with given period are highly likely. The method is demonstrated through a couple of case studies, including a model of the popular Repressilator circuit.

The developmental history of blood cancer begins with mutation acquisition and the resulting malignant clone expansion. The two most prevalent driver mutations found in myeloproliferative neoplasms-JAK2V617F and CALRm-occur in hematopoietic stem cells, which are highly complex to observe in vivo. To circumvent this difficulty, we propose a method relying on mathematical modeling and statistical inference to determine disease initiation and dynamics. Our findings suggest that CALRm mutations tend to occur later in life than JAK2V617F. Our results confirm the higher proliferative advantage of the CALRm malignant clone compared to JAK2V617F. Furthermore, we illustrate how mathematical modeling and Bayesian inference can be used for setting up early screening strategies.

Chemical Reaction Networks (CRN) constitute a formalism used to model biological processes. When the population number is not significant and the system is well-stirred, a Continuous-Time Markov Chain describes its stochastic dynamics. This class of model is characterised by the memoryless property: the future state of the system only depends on the current state.Statistical inference of such CTMCs is complex: likelihood computations are generally intractable. Approximate Bayesian Computation is a recent class of likelihood-free methods for Bayesian inference that allows approximating the posterior distribution with Monte Carlo simulations. It has proven its efficiency in the case of CTMCs.Model-checking was initially developed for assessing hardware and software systems’ reliability. There is a growing interest in the verification of models from Systems Biology to understand the complex molecular interactions within a biological system. Unfortunately, the state space of a CTMC modelled by a CRN quickly explodes or is infinite, which renders its complete exploration infeasible in practice. Statistical model checking methods have been developed to overcome this issue. They simulate the model and compute the ratio of simulations that fulfils a property. Recently, Hybrid Automata Stochastic Logic (HASL) has been introduced for the statistical verification of stochastic models. This temporal logic inherently adopts the statistical point of view of model checking.In this thesis, we focus on statistical inference and verification of CTMCs defined by CRNs. Our main contribution consists in the new formulation of an Approximate Bayesian Computation procedure combined with HASL called automaton-ABC. We apply this high-level method on several tasks of statistical inference and verification for biological CTMCs, including oscillatory models and time-bounded reachability problems. The implementation of our algorithms is documented and has led to a package in the Julia Programming language.

We present an adaptation of the Approximate Bayesian Computation method to estimate the satisfaction probability function of a temporal logic property for Markov Population Models. In this paper, we tackle the problem of estimating the satisfaction probability function of a temporal logic property w.r.t. a parametric Markovian model of Chemical Reaction Network. We want to assess the probability with which the trajectories generated by a parametric Markov Population Model (MPM) satisfy a logical formula over the whole parameter space. In the first step of the work, we formally define a distance between a trajectory of an MPM and a logical property. If the distance is 0, the trajectory satisfies the property. The larger the distance is, further the trajectory is from satisfying the property. In the second step, we adapt the Approximate Bayesian Computation method using the distance defined in the first step. This adaptation yields a new algorithm, called automaton-ABC, whose output is a density function that directly leads to the estimation of the desired satisfaction probability function. We apply our methodology to several examples and models, and we compare it to state-of-the-art techniques. We show that the sequential version of our algorithm relying on ABC-SMC leads to an efficient exploration of the parameter space with respect to the formula and gives good approximations of the satisfaction probability function at a reduced computational cost.

Clear cell renal cell carcinoma (ccRCC) constitutes the most common renal cell carcinoma subtype and has long been recognized as an immunogenic cancer. As such, significant attention has been directed toward optimizing immune-checkpoints (IC)-based therapies. Despite proven benefits, a substantial number of patients remain unresponsive to treatment, suggesting that yet unreported, immunosuppressive mechanisms coexist within tumors and their microenvironment. Here, we comprehensively analyzed and ranked forty-four immune-checkpoints expressed in ccRCC on the basis of in‐depth analysis of RNAseq data collected from the TCGA database and advanced statistical methods designed to obtain the group of checkpoints that best discriminates tumor from healthy tissues. Immunohistochemistry and flow cytometry confirmed and enlarged the bioinformatics results. In particular, by using the recursive feature elimination method, we show that HLA-G, B7H3, PDL-1 and ILT2 are the most relevant genes that characterize ccRCC. Notably, ILT2 expression was detected for the first time on tumor cells. The levels of other ligand-receptor pairs such as CD70:CD27; 4-1BB:4-1BBL; CD40:CD40L; CD86:CTLA4; MHC-II:Lag3; CD200:CD200R; CD244:CD48 were also found highly expressed in tumors compared to adjacent non-tumor tissues. Collectively, our approach provides a comprehensible classification of forty-four IC expressed in ccRCC, some of which were never reported before to be co-expressed in ccRCC. In addition, the algorithms used allowed identifying the most relevant group that best discriminates tumor from healthy tissues. The data can potentially assist on the choice of valuable immune-therapy targets which hold potential for the development of more effective anti-tumor treatments.

Time-bounded reachability problems are concerned with assessing whether a model’s trajectories traverse a given region of the state-space within given time-bounds. In the case of stochastic models reachability is associated with a measure of probability which depends on the model’s parameters. In this paper we propose a methodology that, given a reachability specification (for a parametric stochastic model), allows for computing a reachability related probability distribution on the parameter space, i.e. a distribution that allows for identifying regions of the parameter space for which there is a non-null probability to match the considered reachability specification. The methodology relies on the characterisation of distance between a model’s trajectory and a reachability specification which we show being assessable by using a hybrid automaton as a monitor of a model’s trajectory. An automata-based adaptation of the Approximated Bayesian Computation method is then introduced to estimate the reachability distribution on the parameter space.