Ramgopal R. Mettu

CD4+ T-cell epitope prediction using antigen processing constraints.

T-cell CD4+ epitopes are important targets of immunity against infectious diseases and cancer. State-of-the-art methods for MHC class II epitope prediction rely on supervised learning methods in which an implicit or explicit model of sequence specificity is constructed using a training set of peptides with experimentally tested MHC class II binding affinity. In this paper we present a novel method for CD4+ T-cell eptitope prediction based on modeling antigen-processing constraints. Previous work indicates that dominant CD4+ T-cell epitopes tend to occur adjacent to sites of initial proteolytic cleavage. Given an antigen with known three-dimensional structure, our algorithm first aggregates four types of conformational stability data in order to construct a profile of stability that allows us to identify regions of the protein that are most accessible to proteolysis. Using this profile, we then construct a profile of epitope likelihood based on the pattern of transitions from unstable to stable regions. We validate our method using 35 datasets of experimentally measured CD4+ T cell responses of mice bearing I-Ab or HLA-DR4 alleles as well as of human subjects. Overall, our results show that antigen processing constraints provide a significant source of predictive power. For epitope prediction in single-allele systems, our approach can be combined with sequence-based methods, or used in instances where little or no training data is available. In multiple-allele systems, sequence-based methods can only be used if the allele distribution of a population is known. In contrast, our approach does not make use of MHC binding prediction, and is thus agnostic to MHC class II genotypes.

VHub: Single-stage virtual network mapping through hub location

Abstract Network virtualization allows multiple networks with different protocol stacks to share the same physical infrastructure. A key problem for virtual network providers is the need to efficiently allocate their customers’ virtual network requests to the underlying network infrastructure. This problem is known to be computationally intractable and heuristic solutions continue to be developed. Most existing heuristics use a two-stage approach in which virtual nodes are first placed on physical nodes and virtual links are subsequently mapped. In this paper, we present a novel approach to virtual network mapping that simultaneously maps virtual nodes and links onto the network infrastructure. Our VHub technique formulates the problem of mapping a virtual network request as a mixed integer program that is based on the p-hub median problem. Results from extensive simulations with synthetic and real virtual network requests show that our solution outperforms existing heuristics, including subgraph isomorphism backtracking search. Our approach requires fewer physical resources to accommodate virtual networks and is able to balance load more evenly across the network infrastructure.

Beyond layers: A 3D-aware toolpath algorithm for fused filament fabrication

Fused filament fabrication (FFF) is gaining traction for rapid prototyping and custom fabrication. Existing toolpath generation methods for FFF printers take as input a three-dimensional model of the target object and construct a layered toolpath that will fabricate the object in 2D slices of a chosen thickness. While this approach is computationally straightforward, it can produce toolpaths that can contain significant, yet unnecessary, extrusionless travel. In this paper we propose a novel 3D toolpath generation paradigm that leverages local feature independence in the target object. In contrast to existing FFF slicing methods which print an object layer by layer, our algorithm provides a means to print local features of an object without being constrained to a single layer. The key benefit of our approach is a tremendous reduction in “extrusionless travel,” in which the printer must move between features without performing any extrusion. We show on a benchmark of 409 objects that our method can yield substantial savings in extrusionless travel, 34% on average, that can directly translate to a reduction in total manufacturing time.

Dec-MCTS: Decentralized planning for multi-robot active perception

We propose a decentralized variant of Monte Carlo tree search (MCTS) that is suitable for a variety of tasks in multi-robot active perception. Our algorithm allows each robot to optimize its own actions by maintaining a probability distribution over plans in the joint-action space. Robots periodically communicate a compressed form of their search trees, which are used to update the joint distribution using a distributed optimization approach inspired by variational methods. Our method admits any objective function defined over robot action sequences, assumes intermittent communication, is anytime, and is suitable for online replanning. Our algorithm features a new MCTS tree expansion policy that is designed for our planning scenario. We extend the theoretical analysis of standard MCTS to provide guarantees for convergence rates to the optimal payoff sequence. We evaluate the performance of our method for generalized team orienteering and online active object recognition using real data, and show that it compares favorably to centralized MCTS even with severely degraded communication. These examples demonstrate the suitability of our algorithm for real-world active perception with multiple robots.

A Confidence Measure for Model Fitting with X-Ray Crystallography Data

Structure determination from X-ray crystallography requires numerous stages of iterative refinement between real and reciprocal space. Current methods that fit a model structure to X-ray data therefore utilize a refined experimental electron density map along with a scoring function that characterizes the fit of the density map to structure. Additional information (e.g., from an energy function or conformational statistics) may supplement this score. In this paper, we derive a novel confidence measure for fitting model fragments into X-ray crystallography data. Given any set of conformations under consideration (e.g., a set of sidechain rotamers, or backbone fragments), and a scoring function for those conformations (e.g., least squares fit of the associated model density maps), we give a general-purpose method for assessing the confidence of the best-fit model. For the commonly used least-squares measure of fit, our method analyzes the statistics of the matching scores and estimates the probability that the best-fit conformation is the correct underlying model. To our knowledge, ours is the first method for computing such a confidence measure. To demonstrate the practical utility of our method, we study the problem of sidechain placement and show that our confidence measure can be used to detect and correct incorrect conformational predictions. Over nine proteins with density maps of varying resolutions, the Pearson correlation between predictive accuracy (of least-squares fit) and our confidence measure is quite high, about .89. We show that our approach can guide the use of stereochemical restraints when confidence is low in predictions. We also propose a Bayesian data fusion scheme that integrates our confidence measure to weight the contributon of each source of data, which could potentially be used for combining experimental, modeling, and empirical data in automated structure determination.

Multi-vehicle refill scheduling with queueing

Abstract We consider the problem of refill scheduling for a team of vehicles or robots that must contend for access to a single physical location for refilling. The objective is to minimise time spent in travelling to/from the refill station, and also time lost to queuing (waiting for access). In this paper, we present principled results for this problem in the context of agricultural operations. We first establish that the problem is NP-hard and prove that the maximum number of vehicles that can usefully work together is bounded. We then focus on the design of practical algorithms and present two solutions. The first is an exact algorithm based on dynamic programming that is suitable for small problem instances. The second is an approximate anytime algorithm based on the branch and bound approach that is suitable for large problem instances with many robots. We present simulated results of our algorithms for three classes of agricultural work that cover a range of operations: spot spraying, broadcast spraying and slurry application. We show that the algorithm is reasonably robust to inaccurate prediction of resource utilisation rate, which is difficult to estimate in cases such as spot application of herbicide for weed control, and validate its performance in simulation using realistic scenarios with up to 30 robots.

An improved toolpath generation algorithm for fused filament fabrication

Widely-used methods for toolpath planning in fused filament fabrication slice the input 3D model into 2D layers and construct a toolpath. In prior work (ICRA 2016) we gave a simple greedy algorithm that changed this paradigm and constructed the toolpath in 3D by printing local features in their entirety. This algorithm significantly improved upon layer-based methods, achieving a 34% mean reduction of wasted motion. In this paper we give a new algorithm that is more robust and achieves significantly better performance than the greedy approach. Our algorithm utilizes local search and nearly doubles our prior improvement, achieving a mean/median reduction of 62% over layer-based methods on the same benchmark of over 400 models. We also study toolpath optimality using a novel integer linear programming formulation. This formulation allows us to solve a linear programming relaxation that, while computationally intensive, can give us a lower bound on the optimal solution quality, giving us the ability to rigorously characterize solution quality for a given input model.