Patient-specific models from inter-patient biological models and clinical records

Abstract

Abstract-One of the main goals of systems biology models in a health-care context is to individualise models in order to compute patient-specific predictions for the time evolution of species (e.g., hormones) concentrations. In this paper we present a statistical model checking based approach that, given an interpatient model and a few clinical measurements, computes a value for the model parameter vector (model individualisation) that, with high confidence, is a global minimum for the function evaluating the mismatch between the model predictions and the available measurements. We evaluate effectiveness of the proposed approach by presenting experimental results on using the GynCycle model (describing the feedback mechanisms regulating a number of reproductive hormones) to compute patientspecific predictions for the time evolution of blood concentrations of E2 (Estradiol), P4 (Progesterone), FSH (Follicle-Stimulating Hormone) and LH (Luteinizing Hormone) after a certain number of clinical measurements.

Computing Personalised Treatments through In Silico Clinical Trials. A Case Study on Downregulation in Assisted Reproduction.

“…To do this, we exploit a large VPH model of the Hypothalamic-Pituitary-Gonadal (HPG) axis [36] in order to conduct a multi-arm ISCT. This model has been used as a case study in [39,31], where a parallel algorithm based on statistical model checking techniques and deployed on a High Performance Computing (HPC) infrastructure has been described, which computes a representative set of Virtual Phenotypes (VPs) for the VPH model given as input.…”

“…Here we select, from the population of VPs computed in [39,31] for the HPG axis model [36], a representative subset of 98 different VPs of our HPG axis model, and conduct a distinct arm of our ISCT for each of such phenotypes, in order to compute the lightest (in terms of overall amount of administered drug) treatment still effective for that phenotype.…”

“…In [39,31] a statistical model checking-based approach using a non-uniform sampling process finely tuned against the model has been discussed, and a representative population of VPs for GynCycle consisting of around 10 4 individuals has been computed (once and for all ) from a huge search space of 10 75 possible candidates.…”

“…In order to carry out our multi-arm ISCT, we clustered such VPs by merging in the same cluster phenotypes behaving similarly, where similarity has been defined (along the lines of [39,31]) by means of signal processing metrics. For each obtained cluster, a single representative has been randomly selected.…”

An Integrative Approach for Model Driven Computation of Treatments in Reproductive Medicine

“…The computation of an appropriately large set of biologically admissible (BA) parameters may need several days even on a cluster with many CPUs. Details of the approach can be found elsewhere [12]. The biologically admissible parametrizations may be interpreted also as virtual patients.…”