Predicting age and gender from network telemetry: Implications for privacy and impact on policy.

The systematic monitoring of private communications through the use of information technology pervades the digital age. One result of this is the potential availability of vast amount of data tracking the characteristics of mobile network users. Such data is becoming increasingly accessible for commercial use, while the accessibility of such data raises questions about the degree to which personal information can be protected. Existing regulations may require the removal of personally-identifiable information (PII) from datasets before they can be processed, but research now suggests that powerful machine learning classification methods are capable of targeting individuals for personalized marketing purposes, even in the absence of PII. This study aims to demonstrate how machine learning methods can be deployed to extract demographic characteristics. Specifically, we investigate whether key demographics-gender and age-of mobile users can be accurately identified by third parties using deep learning techniques based solely on observations of the user's interactions within the network. Using an anonymized dataset from a Latin American country, we show the relative ease by which PII in terms of the age and gender demographics can be inferred; specifically, our neural networks model generates an estimate for gender with an accuracy rate of 67%, outperforming decision tree, random forest, and gradient boosting models by a significant margin. Neural networks achieve an even higher accuracy rate of 78% in predicting the subscriber age. These results suggest the need for a more robust regulatory framework governing the collection of personal data to safeguard users from predatory practices motivated by fraudulent intentions, prejudices, or consumer manipulation. We discuss in particular how advances in machine learning have chiseled away a number of General Data Protection Regulation (GDPR) articles designed to protect consumers from the imminent threat of privacy violations.

The dynamic evolution of the power exponent in a universal growth model of tumors.

We have previously reported that a universal growth law, as proposed by West and collaborators for all living organisms, appears to be able to describe also the growth of tumors in vivo after an initial exponential growth phase. In contrast to the assumption of a fixed power exponent p (assumed by West et al. to be equal to 3/4), we propose in this paper a dynamic evolution of p, using experimental data from the cancer literature. In analogy with results obtained by applying scaling laws to the study of fragmentation of solids, the dynamic behaviour of p is related to the evolution of the fractal topology of neoplastic vascular systems. Our model might be applied for diagnostic purposes to mark the emergence of an efficient neo-angiogenetic structure if the results of our in silico experiments are confirmed by clinical observations.

Evolutionary game theory in an agent-based brain tumor model: exploring the 'Genotype-Phenotype' link.

To investigate the genotype-phenotype link in a polyclonal cancer cell population, here we introduce evolutionary game theory into our previously developed agent-based brain tumor model. We model the heterogeneous cell population as a mixture of two distinct genotypes: the more proliferative Type A and the more migratory Type B. Our agent-based simulations reveal a phase transition in the tumor's velocity of spatial expansion linking the tumor fitness to genotypic composition. Specifically, velocity initially falls as rising payoffs reward the interactions among the more stationary Type A cells, but unexpectedly accelerates again when these A-A payoffs increase even further. At this latter accelerating stage, fewer migratory Type B cells appear to confer a competitive advantage in terms of the tumor's spatial aggression over the overall numerically dominating Type A cells, which in turn leads to an acceleration of the overall tumor dynamics while its surface roughness declines. We discuss potential implications of our findings for cancer research.

Correlating velocity patterns with spatial dynamics in glioma cell migration.

Highly malignant neuroepithelial tumors are known for their extensive tissue invasion. Investigating the relationship between their spatial behavior and temporal patterns by employing detrended fluctuation analysis (DFA), we report here that faster glioma cell motility is accompanied by both greater predictability of the cells' migration velocity and concomitantly, more directionality in the cells' migration paths. Implications of this finding for both experimental and clinical cancer research are discussed.

Insights from a novel tumor model: Indications for a quantitative link between tumor growth and invasion.

Both the lack of nutrient supply and rising mechanical stress exerted by the microenvironment appear to be able to cause discrepancies between the actual, observed tumor mass and that predicted by West et al.'s [A general model for ontogenetic growth. Nature 2001;413:628-31] universal growth model. Using our previously developed model we hypothesize here, that (1) solid tumor growth and cell invasion are linked, not only qualitatively but also quantitatively, that (2) the onset of invasion marks the time point when the tumor's cell density reaches a compaction maximum, and that (3) tumor cell invasion, reduction of mechanical confinement and angiogenesis can all contribute to an increase in the actual tumor mass m towards the level m(W) predicted by West et al.'s universal growth curve. These novel insights contribute to our understanding of tumorigenesis and thus may have important implications not only for experimental cancer research but also be of value for clinical purposes such as for predictions of tumor growth dynamics and treatment impact.

Simulating the impact of a molecular 'decision-process' on cellular phenotype and multicellular patterns in brain tumors.

Experimental evidence indicates that human brain cancer cells proliferate or migrate, yet do not display both phenotypes at the same time. Here, we present a novel computational model simulating this cellular decision-process leading up to either phenotype based on a molecular interaction network of genes and proteins. The model's regulatory network consists of the epidermal growth factor receptor (EGFR), its ligand transforming growth factor-alpha (TGF alpha), the downstream enzyme phospholipaseC-gamma (PLC gamma) and a mitosis-associated response pathway. This network is activated by autocrine TGF alpha secretion, and the EGFR-dependent downstream signaling this step triggers, as well as modulated by an extrinsic nutritive glucose gradient. Employing a framework of mass action kinetics within a multiscale agent-based environment, we analyse both the emergent multicellular behavior of tumor growth and the single-cell molecular profiles that change over time and space. Our results show that one can indeed simulate the dichotomy between cell migration and proliferation based solely on an EGFR decision network. It turns out that these behavioral decisions on the single cell level impact the spatial dynamics of the entire cancerous system. Furthermore, the simulation results yield intriguing experimentally testable hypotheses also on the sub-cellular level such as spatial cytosolic polarization of PLC gamma towards an extrinsic chemotactic gradient. Implications of these results for future works, both on the modeling and experimental side are discussed.

The impact of "search precision" in an agent-based tumor model.

Several cell surface receptors are overexpressed in malignant brain tumors and reportedly involved in tumor progression and invasion. It is unclear, however, whether such an improvement of cellular signal reception leads to a monotonic increase in the tumor system's average velocity during invasion or whether there is a maximum threshold beyond which the average velocity starts to decelerate. To gain more insight into the systemic effects of such cellular search precision modulations, this study proposes a two-dimensional agent-based model in which the spatio-temporal expansion of malignant brain tumor cells is guided by environmental heterogeneities in mechanical confinement, toxic metabolites and nutrient sources. Here, the spatial field of action is represented by an adaptive grid lattice, which corresponds to the experimental finding that tumor cells are more likely to follow each other along preformed pathways. Another prominent feature is the dual threshold concept for both nutrient level and toxicity, which determine whether cells proliferate, migrate, remain quiescent or die in the next period. The numerical results from varying the key parameters encoding the capability of tumor cells to invade and their ability to proliferate indicate an emergent behavior. Specifically, increasing invasiveness not only leads to an increase in maximum expansion velocity, but also requires a more precise spatial search process, corresponding to an improved cell signal reception, in order to obtain maximum velocity. To increase cellular invasiveness beyond the maximum that can be achieved by exclusively tuning the motility parameter, it requires an additional reduction in the cells' proliferation rate and prompts an even more biased search process. Most interestingly, however, a prominent phase transition suggests that tumor cells do not employ a 100 percent search precision to attain maximum spatial velocity. These findings argue for a selection advantage conferred by limited randomness in processing spatial search and indicate that our computational platform may prove valuable in investigating emergent, multicellular tumor patterns caused by alterations on the molecular level.

Emerging patterns in tumor systems: simulating the dynamics of multicellular clusters with an agent-based spatial agglomeration model.

Brain cancer cells invade early on surrounding parenchyma, which makes it impossible to surgically remove all tumor cells and thus significantly worsens the prognosis of the patient. Specific structural elements such as multicellular clusters have been seen in experimental settings to emerge within the invasive cell system and are believed to express the systems' guidance toward nutritive sites in a heterogeneous environment. Based on these observations, we developed a novel agent-based model of spatio-temporal search and agglomeration to investigate the dynamics of cell motility and aggregation with the assumption that tumors behave as complex dynamic self-organizing biosystems. In this model, virtual cells migrate because they are attracted by higher nutrient concentrations and to avoid overpopulated areas with high levels of toxic metabolites. A specific feature of our model is the capability of cells to search both globally and locally. This concept is applied to simulate cell-surface receptor-mediated information processing of tumor cells such that a cell searching for a more growth-permissive place "learns" the information content of a brain tissue region within a two-dimensional lattice in two stages, processing first the global and then the local input. In both stages, differences in microenvironment characteristics define distinctions in energy expenditure for a moving cell and thus influence cell migration, proliferation, agglomeration, and cell death. Numerical results of our model show a phase transition leading to the emergence of two distinct spatio-temporal patterns depending on the dominant search mechanism. If global search is dominant, the result is a small number of large clusters exhibiting rapid spatial expansion but shorter lifetime of the tumor system. By contrast, if local search is dominant, the trade-off is many small clusters with longer lifetime but much slower velocity of expansion. Furthermore, in the case of such dominant local search, the model reveals an expansive advantage for tumor cell populations with a lower nutrient-depletion rate. Important implications of these results for cancer research are discussed.