ABSTRACT. The emergence of drug-resistance is a major challenge in chemotherapy. In this talk we will overview some of our recent mathematical models for describing the dynamics of drug-resistance in solid tumors. These models follow the dynamics of the tumor, assuming that the cancer cell population depends on a phenotype variable that corresponds to the resistance level to a cytotoxic drug. Under certain conditions, our models predict that multiple resistant traits emerge at different locations within the tumor, corresponding to heterogeneous tumors. We show that a higher drug dosage may delay a relapse, yet, when this happens, a more resistant trait emerges. We will show how mathematics can be used to propose an efficient drug schedule aiming at minimizing the growth rate of the most resistant trait, and how such resistant cells can be eliminated.

ABSTRACT. Breast cancer is the most common cancer among women both in developed and developing countries. Early detection and diagnosis of breast cancer may reduce its mortality and improve the quality of life. Computer-aided detection (CADx) and computer-aided diagnosis (CAD) techniques have shown promise for reducing the burden of human expert reading and improve the accuracy and reproducibility of results. Sparse analysis techniques have produced relevant results for representing and recognizing imaging patterns. In this work we propose a method for Label Consistent Spatially Localized Ensemble Sparse Analysis (LC-SLESA). In this work we apply dictionary learning to our block based sparse analysis method to classify breast lesions as benign or malignant. The performance of our method and our method in conjunction with LC-KSVD dictionary learning is evaluated using 10-, 20-, and 30-fold cross validation on the MIAS dataset. Our results indicate that the proposed sparse analyses may be a useful component for breast cancer screening applications.

ABSTRACT. Oral cancer is most prevalent in low- and middle-income countries where it is associated with late diagnosis. A significant factor for this is the limited access to specialist diagnosis. The use of artificial intelligence for decision making on oral cavity images has the potential to improve cancer management and survival rates. This study forms part of the MeMoSA® (Mobile Mouth Screening Anywhere) project. In this paper, we extended on our previous deep learning work and focused on the binary image classification of ‘referral’ vs. ‘non-referral’. Transfer learning was applied, with several common pre-trained deep convolutional neural network architectures compared for the task of fine-tuning to a small oral image dataset. Improvements to our previous work were made, with an accuracy of 80.88% achieved and a corresponding sensitivity of 85.71% and specificity of 76.42%.

ABSTRACT. Mathematical and computational oncology has increased the pace of cancer research towards the advancement of personalized therapy. Serving the pressing need to exploit the currently underutilized data, such approaches bring a significant clinical advantage in tailoring the therapy. CHIMERA is a novel system that combines mechanistic modelling and machine learning for personalized chemotherapy and surgery sequencing in breast cancer. It optimizes decision-making in personalized breast cancer therapy by connecting tumor growth behaviour and chemotherapy effects through predictive modelling and learning. We demonstrate the capabilities of CHIMERA in learning simultaneously the tumor growth patterns, across several types of breast cancer, and the pharmacokinetics of a typical breast cancer chemotoxic drug. The learnt functions are subsequently used to predict how to sequence the intervention. We demonstrate the versatility of CHIMERA in simultaneously learning tumor growth and pharmacokinetics under two, typically used, chemotherapy protocol hypotheses.

ABSTRACT. It is the well-accepted concept that tumors evolve under the pressure of immune responses and escape from them. Immune checkpoint inhibitors (ICIs) are expected to reactivate antitumor immunity and inhibit tumor progression. However, the durable benefits by ICI treatments are limited to the minority of patients. Therefore, it is important to study the mechanisms of tumor evolution interacting with immune cells and to reveal the condition that ICIs become effective. Here, we developed a mathematical model of the tumor evolution in which the immune responses are changed in accordance with the accumulation of mutations in the tumor cells. We assume that as the mutational load increases, the strength of the immune responses, which slows tumor growth, and the chance of immune escape, which prevents immune responses, simultaneously increase. To consider the variety of patients, we performed computer simulations with randomly determined parameters. Depending on the parameters, some tumors quickly grew and were detected with low mutation load, and some slowly grew and were detected with high mutation load. After a tumor is detected, we simulated the anti-tumor treatment in which the parameters of intrinsic tumor growth and immune responses were modified and checked whether the treatment was effective or not. As a result, we confirmed the known tendency that patients who had high mutational load were likely to have a durable benefit. Furthermore, we suggest two predictions: i) the combination of chemotherapy and ICI therapy would increase the proportion of the responder; and ii) the growth rate of tumor cells may be informative to determine whether a patient is responder or not. These predictions would be confirmed by clinical data.

ABSTRACT. Small cell lung cancer (SCLC) is an aggressive neuroendocrine carcinoma known for rapid metastasis and recurrence following treatment. The standard of care (etoposide + cis-platinum and radiation) has hardly changed in over 30 years, with dismal outcomes for patients. There is a pressing need, therefore, for new therapies to treat this deadly disease. Although previously thought of as a homogeneous disease comprised of “small round blue cells,” recent studies have identified numerous subtypes of SCLC that support tumor growth, treatment evasion, and metastasis. Recent work from our lab identified four distinct SCLC subtypes, three of which were described previously and a fourth which is novel and broadly insensitive to several classes of therapeutic agents, pointing to a possible role in treatment resistance.

To investigate SCLC tumor growth dynamics and identify factors that can modulate tumor composition, we constructed a population dynamics model incorporating cell-cell interactions. The model features three neuroendocrine (NE) subtypes and one non-NE subtype that provides trophic support to NE cells via secreted factors that enhance cell division and inhibit death. Cells can reversibly transition between all NE subtypes but only one NE variant can transition into the non-NE state. The three NE subtypes are also assumed to secrete factors that inhibit division of non-NE cells. We fit the model to tumor data from two genetically engineered mouse models (GEMMs) with distinct subtype compositions to identify driving factors that can modulate tumor composition. Our analysis points to transition rates and cell-cell interactions as crucial for defining and maintaining subtype proportions within tumors. We speculate that Myc may be important in phenotypic transitions and the molecular composition of secreted factors, in line with observed effects of Myc overexpression in GEMMs.

ABSTRACT. We study the performance of a variant detection method that is based on a property of tumor phylogenetic tree. Our major contributions are two folds. First, we show the property of tumor phylogenetic tree: the total patterns of mutations are restricted if a multi-regional mutation profile follows a corresponding tumor phylogenetic tree, where a multi-regional mutation profile is a matrix in which predictions of somatic mutations at the corresponding tumor regions are listed. Second, we evaluate the lower and upper bounds of specificity and sensitivity of a phylogeny-based somatic variant detection method under several situations. In the evaluation, we conduct patient-wise variant detection from a noisy multi-regional mutation profile matrix for some genomic positions by utilizing the phylogenetic property; we assume that the phylogenetic information can be extracted from another mutation profile matrix that contains accurate candidates at different genomic positions from the noisy ones. From the evaluation, we find that higher sensitivity is not guaranteed in the phylogeny-based variant detection, but higher specificity is guaranteed for several cases. These findings indicate the tumor phylogeny gives more merit for variant detection based on erroneous long-read sequencers (e.g. Oxford nanopore sequencers) than that based on accurate short-read sequencers (e.g., Illumina sequencer).

ABSTRACT. Hematopoiesis is a complex process in which stem cells in the bone marrow produce all the cells circulating in the blood. When specific mutations occur in the stem cells, such as the JAK2V617F mutation, for example, this process can be altered leading to the development of hematological malignancies such as Myeloproliferative Neoplasms. Interferon alpha (IFN alpha) is a treatment that allows a hematological response in patients, but also in some cases a molecular response. However, its precise mechanism of action is still poorly understood, and there are no clear guidelines on the use of this treatment for clinicians. In this article, we model the action of the IFN alpha on the hematopoiesis dynamics in order to study in particular its effect at the level of stem cells. Using a Bayesian parametric estimation method and data from a patient under treatment for several years, we show that IFN alpha could act on cancer stem cells by promoting their quiescence exit and thus their proliferation while allowing their exhaustion from the stem cells stock by increasing their proportion to make differentiated divisions.

ABSTRACT. The functions of only a minority of genes in any species is known. And even in those cases the functional annotation is highly incomplete and largely devoid of context. At an even more fundamental level, how can we know whether a gene serves any relevant biological function in a given context? In this informal presentation we will discuss a few vignettes related to the broad questions of context-specific functions of genes, in a variety of contexts from bacterial response to drugs, normal tissues, and cancer.

ABSTRACT. Characterizing the mode – the way, manner, or pattern – of evolution in tumours is important for clinical forecasting and optimizing cancer treatment. DNA sequencing studies have inferred various modes, including branching, punctuated and neutral evolution, but it is unclear why a particular pattern predominates in any given tumour. I will argue that differences in tumour architecture can explain the variety of observed genetic patterns. I will pre-sent results of spatially explicit population genetic models (Fig. 1) showing that, within biologically relevant parameter ranges, human tumours are expected to exhibit four distinct onco-evolutionary modes (oncoevotypes), governed by the mode of cell dispersal and the range of cell-cell interaction. New quantitative indices will be introduced for describing and classifying these oncoevotypes, and for comparing with multi-region sequencing data. I will further present an investigation of when, why and how intratumour heterogeneity can be used to forecast tumour growth rate and progression-free survival. In cohorts of tumours with diverse evolutionary parameters, I will show that clonal diversity is expected to be a reliable predictor of both growth rate and progression-free survival. I will thus provide explanations – grounded in evolutionary theory – for empirical findings in various cancers. This work informs the search for new prognostic biomarkers and con-tributes to the development of predictive oncology.