Many multi-objective optimisation problems incorporate computationally or nancially expensive objective functions. State-of-theart algorithms therefore construct surrogate model(s) of the parameter space to objective functions mapping to guide the choice of the next solution to expensively evaluate. Starting from an initial set of solutions, an in ll criterion-a surrogate-based indicator of quality-is extremised to determine which solution to evaluate next, until the budget of expensive evaluations is exhausted. Many successful in ll criteria are dependent on multi-dimensional integration, which may result in in ll criteria that are themselves impractically expensive. We propose a computationally cheap in ll criterion based on the minimum probability of improvement over the estimated Pareto set. We also present a range of set-based scalarisation methods modelling hypervolume contribution, dominance ratio and distance measures. ese permit the use of straightforward expected improvement as a cheap in ll criterion. We investigated the performance of these novel strategies on standard multi-objective test problems, and compared them with the popular SMS-EGO and ParEGO methods. Unsurprisingly, our experiments show that the best strategy is problem dependent, but in many cases a cheaper strategy is at least as good as more expensive alternatives. CCS CONCEPTS •Computing methodologies → Gaussian processes; Modeling methodologies; •Applied computing → Multi-criterion optimization and decision-making; •Mathematics of computing → Probabilistic algorithms;

The intriguing properties of reduced graphene oxide (rGO) have paved the way for a number of potential biomedical applications such as drug delivery, tissue engineering, gene delivery and bio-sensing. Over the last decade, there have been escalating concerns regarding the possible toxic effects, behaviour and fate of rGO in living systems and environments. This paper reports on integrative chemical-biological interactions of rGO with lung cancer cells, i.e. A549 and SKMES-1, to determine its potential toxicological impacts on them, as a function of its concentration. Cell viability, early and late apoptosis and necrosis were measured to determine oxidative stress potential, and induction of apoptosis for the first time by comparing two lung cancer cells. We also showed the general trend between cell death rates and concentrations for different cell types using a Gaussian process regression model. At low concentrations, rGO was shown to significantly produce late apoptosis and necrosis rather than early apoptotic events, suggesting that it was able to disintegrate the cellular membranes in a dose dependent manner. For the toxicity exposures undertaken, late apoptosis and necrosis occurred, which was most likely resultant from limited bioavailability of unmodified rGO in lung cancer cells.

Overexpression and secretion of the enzymes cathepsin D (CathD) and cathepsin L (CathL) is associated with metastasis in several human cancers. As a superfamily, extracellularly, these proteins may act within the tumor microenvironment to drive cancer progression, proliferation, invasion and metastasis. Therefore, it is important to discover novel therapeutic treatment strategies to target CathD and CathL and potentially impede metastasis. Graphene oxide (GO) could form the basis of such a strategy by acting as an adsorbent for pro-metastatic enzymes. Here, we have conducted research into the potential of targeted anti-metastatic therapy using GO to adsorb these pro-tumorigenic enzymes. Binding of CathD/L to GO revealed that CathD/L were adsorbed onto the surface of GO through its cationic and hydrophilic residues. This work could provide a roadmap for the rational integration of CathD/L-targeting agents into clinical settings.

Wireless sensor networks frequently use multi-path routing schemes between nodes and a base station. Multi-path routing confers additional robustness against link failure, but in batterypowered networks it is desirable to choose paths which maximise the overall network lifetime -the time at which a battery is first exhausted. We introduce multi-objective evolutionary algorithms to find the routings which approximate the optimal trade-off between network lifetime and robustness. A novel measure of network robustness, the fragility, is introduced. We show that the distribution of traffic between paths in a given multi-path scheme that optimises lifetime or fragility may be found by solving the appropriate linear program. A multi-objective evolutionary algorithm is used to solve the combinatorial optimisation problem of choosing routings and traffic distributions that give the optimal trade-off between network lifetime and robustness. Efficiency is achieved by pruning the search space using k-shortest paths, braided and edge disjoint paths. The method is demonstrated on synthetic networks and a real network deployed at the Victoria & Albert Museum, London. For these networks, using only two paths per node, we locate routings with lifetimes within 3% of those obtained with unlimited paths per node. In addition, routings which halve the network fragility are located. We also show that the evolutionary multi-path routing can achieve significant improvement in performance over a braided multi-path scheme.