Abstract Therapeutic antibodies are an important and rapidly growing drug modality. However, the design discovery of early-stage antibody therapeutics remain a time cost-intensive endeavor. Here we present end-to-end Bayesian, language model-based method for designing large diverse libraries high-affinity single-chain variable fragments (scFvs) that then empirically measured. In head-to-head comparison with directed evolution approach, show best scFv generated from our represents 28.7-fold...

The opportunities and challenges of adapting applying AI principles to synbio.

Temporal streams of interactions are commonly aggregated into dynamic networks for temporal analysis. Results this analysis greatly affected by the resolution at which original data aggregated. The mismatch between inherent scale underlying process and that is performed can obscure important insights lead to wrong conclusions. To day, there no established framework choosing appropriate interactions. Our paper offers first step towards formalization problem. We show a general class...

Abstract COVID-19 epidemics have varied dramatically in nature across the United States, where some counties clear peaks infections, and others had a multitude of unpredictable non-distinct peaks. Our lack understanding how pandemic has evolved leads to increasing errors our ability predict spread disease. This work seeks explain this diversity epidemic progressions by considering an extension compartmental SEIRD model. The model we propose uses neural network infection rate as function both...

Meshes are used to represent complex objects in high fidelity physics simulators across a variety of domains, such as radar sensing and aerodynamics. There is growing interest using neural networks accelerate simulations, also body work on applying directly irregular mesh data. Since multiple topologies can the same object, augmentation typically required handle topological variation when training networks. Due sensitivity small changes shape, it challenging use these augmentations...

Applied network science often involves preprocessing data before applying a network-analysis method, and there is typically theoretical disconnect between these steps. For example, it common to aggregate time-varying into windows prior analysis, the tradeoffs of this are not well understood. Focusing on problem detecting small communities in multilayer networks, we study effects layer aggregation by developing random-matrix theory for modularity matrices associated with layer-aggregated...

Topic modeling continues to grow as a popular technique for finding hidden patterns, well grouping collections of new types text and non-text data. Recent years have witnessed growing body work in developing metrics techniques evaluating the quality topic models topics they generate. This is particularly true data where significant attention has been given semantic interpretability using measures such coherence. It shown however that assessments based on coherence do not always align with...

Group living is a life history strategy employed by many organisms. This often difficult to study because the exact boundaries of group can be unclear. Weaverbirds present an ideal model for living, their colonies occupy space with discrete boundaries: single tree. We examined one aspect living. nest placement, in three Kenyan weaverbird species: Black-capped Weaver (Pseudonigrita cabanisi), Grey-capped (P. arnaudi) and White-browed Sparrow (Ploceropasser mahali). asked which environmental,...

The control of complex dynamic networks has many applications ranging from the management electric power grids to regulation biological cellular networks. Prior work focused on understanding relationship between cascading behavior and network structure, role critical nodes, conditions necessary for sustained cascade propagation. Recent begun examine general approaches dynamically influencing processes guide or a targeted subnetwork towards desired state. This paper models an epidemic as...

For any stream of time-stamped edges that form a dynamic network, an important choice is the aggregation granularity analyst uses to bin data. Picking such windowing data often done by hand, or left up technology collecting However, can make big difference in properties network. This time scale detection problem. In previous work, this problem solved with heuristic as unsupervised task. As problem, it difficult measure how well given algorithm performs. addition, we show quality dependent on...

Therapeutic antibody development has become an increasingly popular approach for drug development. To date, therapeutics are largely developed using large scale experimental screens of libraries containing hundreds millions sequences. The high cost and difficulty developing therapeutic antibodies create a pressing need computational methods to predict properties bespoke designs. However, the relationship between sequence activity is complex physical process traditional iterative design...

Word embedding models offer continuous vector representations that can capture rich contextual semantics based on their word co-occurrence patterns. While these vectors provide very effective features used in many NLP tasks such as clustering similar words and inferring learning relationships, challenges open research questions remain. In this paper, we propose a solution aligns variations of the same model (or different models) joint low-dimensional latent space leveraging carefully...

We report on the Second Workshop Mining Networks and Graphs held at 2015 SIAM International Conference Data Mining. This half-day workshop consisted of a keynote talk, four technical paper presentations, one demonstration, panel future challenges in mining large networks. summarize main highlights workshop, including expanded written summaries provided by panelists. The current discussed elaborated here provide valuable guidance for research field

Abstract Complex networks are often either too large for full exploration, partially accessible, or observed. Downstream learning tasks on these incomplete can produce low quality results. In addition, reducing the incompleteness of network be costly and nontrivial. As a result, discovery algorithms optimized specific downstream given resource collection constraints great interest. this paper, we formulate task-specific problem as sequential decision-making problem. Our task is selective...

A common problem in modern graph analysis is the detection of communities, an example which a single anomalously dense subgraph. Recent results have demonstrated fundamental limit for this when using spectral modularity. In paper, we demonstrate implication these on subgraph cue vertex provided, indicating one vertices community interest. Several recent algorithms local are applied context, and compare their empirical performance to that simple method used derive theoretical limits.

A new approach for targeted graph sampling is proposed in which and classification occur together, content-based homophily exploited to achieve improved performance. The application of network discovery relevant content considered using an that may be generalized a broad class vertex properties. resulting procedure provides the initial step analytic processing chain whose performance directly affected by quality sampling. algorithm measured with real data observed on social media site....

Graphs are powerful abstractions for capturing complex relationships in diverse application settings. An active area of research focuses on theoretical models that define the generative mechanism a graph. Yet given complexity and inherent noise real datasets, it is still very challenging to identify best model observed We discuss framework graph selection leverages long list topological properties random forest classifier learn classify different instances. fully characterize discriminative...

Abstract Therapeutic antibodies are an important and rapidly growing drug modality. However, the design discovery of early-stage antibody therapeutics remain a time cost-intensive endeavor. In this work, we present end-to-end Bayesian, language model-based method for designing large diverse libraries high-affinity single-chain variable fragments (scFvs). We integrate target-specific binding affinities with information from millions natural protein sequences in probabilistic machine learning...

In this paper, we present a novel approach based on the random walk process for finding meaningful representations of graph model. Our leverages transient behavior many short walks with initialization mechanisms to generate model discriminative features. These features are able capture more comprehensive structural signature underlying The resulting representation is invariant both node permutation and size graph, allowing direct comparison between large classes graphs. We test our two...