Huey Eng Chua

Computational cell fate modelling for discovery of rewiring in apoptotic network for enhanced cancer drug sensitivity

The ongoing cancer research has shown that malignant tumour cells have highly disrupted signalling transduction pathways. In cancer cells, signalling pathways are altered to satisfy the demands of continuous proliferation and survival. The changes in signalling pathways supporting uncontrolled cell growth, termed as rewiring, can lead to dysregulation of cell fates e.g. apoptosis. Hence comparative analysis of normal and oncogenic signal transduction pathways may provide insights into mechanisms of cancer drug-resistance and facilitate the discovery of novel and effective anti-cancer therapies. Here we propose a hybrid modelling approach based on ordinary differential equation (ODE) and machine learning to map network rewiring in the apoptotic pathways that may be responsible for the increase of drug sensitivity of tumour cells in triple-negative breast cancer. Our method employs Genetic Algorithm to search for the most likely network topologies by iteratively generating simulated protein phosphorylation data using ODEs and the rewired network and then fitting the simulated data with real data of cancer signalling and cell fate. Most of our predictions are consistent with experimental evidence from literature. Combining the strengths of knowledge-driven and data-driven approaches, our hybrid model can help uncover molecular mechanisms of cancer cell fate at systems level.

TENET: topological feature-based target characterization in signalling networks

MOTIVATION Target characterization for a biochemical network is a heuristic evaluation process that produces a characterization model that may aid in predicting the suitability of each molecule for drug targeting. These approaches are typically used in drug research to identify novel potential targets using insights from known targets. Traditional approaches that characterize targets based on their molecular characteristics and biological function require extensive experimental study of each protein and are infeasible for evaluating larger networks with poorly understood proteins. Moreover, they fail to exploit network connectivity information which is now available from systems biology methods. Adopting a network-based approach by characterizing targets using network features provides greater insights that complement these traditional techniques. To this end, we present Tenet (Target charactErization using NEtwork Topology), a network-based approach that characterizes known targets in signalling networks using topological features. RESULTS Tenet first computes a set of topological features and then leverages a support vector machine-based approach to identify predictive topological features that characterizes known targets. A characterization model is generated and it specifies which topological features are important for discriminating the targets and how these features should be combined to quantify the likelihood of a node being a target. We empirically study the performance of Tenet from a wide variety of aspects, using several signalling networks from BioModels with real-world curated outcomes. Results demonstrate its effectiveness and superiority in comparison to state-of-the-art approaches. AVAILABILITY AND IMPLEMENTATION Our software is available freely for non-commercial purposes from: https://sites.google.com/site/cosbyntu/softwares/tenet CONTACT hechua@ntu.edu.sg or assourav@ntu.edu.sg SUPPLEMENTARY INFORMATION Supplementary data are available at Bioinformatics online.

ViSual: An HCI-inspired simulator for blending visual subgraph query construction and processing

In [3], we laid out the vision of a novel graph query processing paradigm, where visual subgraph query formulation is interleaved (or “blended”) with query processing by exploiting the latency offered by the gui. Our recent attempts at implementing this vision [6], [7] do not provide any robust framework to systematically investigate the performance of this novel paradigm. This is because it is prohibitively expensive to engage a large number of users to formulate a large number of visual queries in order to measure the performance of blending query formulation with query processing. In this demonstration, we present a novel synthetic visual subgraph query simulator called ViSual that can evaluate the performance of this paradigm for a large number of visual subgraph queries without requiring a large number of users to formulate them. Specifically, it leverages principles from hci to quantify the gui latency that is necessary to realistically simulate blending of query formulation and query processing.

One feature doesn't fit all: characterizing topological features of targets in signaling networks

A key challenge facing drug discovery is the identification of target(s) in a signaling network whose perturbation results in a desired therapeutic outcome. Recent studies have shown that analysis of biological networks based on topology can facilitate target identification by providing valuable information on characteristics of targets. In this paper, we present an algorithm called Differ that discovers the discriminative topological features (dtf) from a signaling network to distinguish the targets from the non-targets. Our empirical study on five signaling networks reveals that the majority of dtfs are able to identify most of the known targets in these networks. Furthermore, they are distinct for different networks. That is, no single topological feature can characterise targets in all signaling networks. This is in contrast to the findings in [28] where bridging nodes are considered to be good targets with low lethality across several ppi networks.

STEROID: in silico heuristic target combination identification for disease-related signaling networks

Given a signaling network, the target combination identification problem aims to predict efficacious and safe target combinations for treatment of a disease. State-of-the-art in silico methods use Monte Carlo simulated annealing (mcsa) to modify a candidate solution stochastically, and use the Metropolis criterion to accept or reject the proposed modifications. However, such stochastic modifications ignore the impact of the choice of targets and their activities on the combinations therapeutic effect and off-target effects which directly affect the solution quality. In this paper, we present Steroid, a novel method that addresses this limitation by leveraging two additional heuristic criteria to minimize off-target effects and achieve synergy for candidate modification. Specifically, off-target effects measure the unintended response of a signaling network to the target combination and is generally associated with toxicity. Synergy occurs when a pair of targets exerts effects that are greater than the sum of their individual effects, and is generally a beneficial strategy for maximizing effect while minimizing toxicity. Our empirical study on the cancer-related mapk-pi3k network demonstrates the superiority of Steroid in comparison to mcsa-based approaches. Specifically, Steroid is an order of magnitude faster and yet yields biologically relevant synergistic target combinations with significantly lower off-target effects.

BOOMER: Blending Visual Formulation and Processing of P -Homomorphic Queries on Large Networks

Visual graph query interfaces (a.k.a GUI) make it easy for non-expert users to query graphs. Recent research has laid out and implemented a vision of a novel subgraph query processing paradigm where the latency offered by the GUI is exploited to blend visual query construction and processing by generating and refining candidate result matches iteratively during query formulation. This paradigm brings in several potential benefits such as superior system response time (srt) and opportunities to enhance usability of graph databases. However, these early efforts focused on subgraph isomorphism-based graph queries where blending is performed by iterative edge-to-edge mapping. In this paper, we explore how this vision can be realized for more generic but complex 1-1 p-homomorphic p-hom) queries introduced by Fan et al. A 1-1 p-hom query maps an edge of the query to paths in the data graph. We present a novel framework called BOOMER for blending bounded 1-1 p-hom (bph ) queries, a variant of 1-1 p-hom where the length of the path is bounded instead of arbitrary length. Our framework is based on a novel online , adaptive indexing scheme called cap index. We present two strategies for CAP index construction, immediate and deferment-based, and show how they can be utilized to facilitate judicious interleaving of visual bph query formulation and query processing. BOOMER is also amenable to modifications to a bph query during visual formulation. Experiments on real-world datasets demonstrate both efficiency and effectiveness of Boomer for realizing the visual querying paradigm on an important type of graph query.

Synergistic target combination prediction from curated signaling networks: Machine learning meets systems biology and pharmacology

Given a signaling network, the target combination prediction problem aims to predict efficacious and safe target combinations for combination therapy. State-of-the-art in silico methods use Monte Carlo simulated annealing (mcsa) to modify a candidate solution stochastically, and use the Metropolis criterion to accept or reject the proposed modifications. However, such stochastic modifications ignore the impact of the choice of targets and their activities on the combinations therapeutic effect and off-target effects, which directly affect the solution quality. In this paper, we present mascot, a method that addresses this limitation by leveraging two additional heuristic criteria to minimize off-target effects and achieve synergy for candidate modification. Specifically, off-target effects measure the unintended response of a signaling network to the target combination and is often associated with toxicity. Synergy occurs when a pair of targets exerts effects that are greater than the sum of their individual effects, and is generally a beneficial strategy for maximizing effect while minimizing toxicity. mascot leverages on a machine learning-based target prioritization method which prioritizes potential targets in a given disease-associated network to select more effective targets (better therapeutic effect and/or lower off-target effects); and on Loewe additivity theory from pharmacology which assesses the non-additive effects in a combination drug treatment to select synergistic target activities. Our experimental study on two disease-related signaling networks demonstrates the superiority of mascot in comparison to existing approaches.

TAPESTRY: Network-centric Target Prioritization in Disease-related Signaling Networks

Target prioritization ranks molecules in biological networks according to a score that seeks to identify molecules that fulfill particular roles (e.g., drug targets). We study this problem in the context of partial information (e.g., unknown targets) and present TAPESTRY, a network-based approach that prioritizes candidate targets in a given signaling network with unknown targets by utilizing knowledge (target characteristics) gained from curated targets in another set of signaling networks. We consider both topological and dynamic features and use a weighted sum approach to examine the relative influence of these two classes of features on the prioritization results. TAPESTRY exploits a knowledge base of characterization models and predictive topological features of a set of signaling networks (candidate networks) with curated targets. Then, given a signaling network G with unknown targets, TAPESTRY identifies a candidate network most similar to G and selects its characterization model as prioritization model for computing a topological feature-based rank of each candidate node in G. Next, a dynamic feature-based rank is computed for these nodes by leveraging the time-series curves of ODEs associated with the edges in G. Finally, these two ranks are integrated and used for prioritizing candidate targets. We experimentally study the performance of TAPESTRY using signaling networks from BioModels with real-world curated outcomes. Our results demonstrate its effectiveness and superiority in comparison to state-of-the-art approaches.

TROVE: a user-friendly tool for visualizing and analyzing cancer hallmarks in signaling networks

Summary Cancer hallmarks, a concept that seeks to explain the complexity of cancer initiation and development, provide a new perspective of studying cancer signaling which could lead to a greater understanding of this complex disease. However, to the best of our knowledge, there is currently a lack of tools that support such hallmark-based study of the cancer signaling network, thereby impeding the gain of knowledge in this area. We present TROVE, a user-friendly software that facilitates hallmark annotation, visualization and analysis in cancer signaling networks. In particular, TROVE facilitates hallmark analysis specific to particular cancer types. Availability and Implementation Available under the Eclipse Public License from: https://sites.google.com/site/cosbyntu/softwares/trove and https://github.com/trove2017/Trove. Contact hechua@ntu.edu.sg or assourav@ntu.edu.sg.

FacetsViewer: A Tool for Multi-faceted Decomposition of Complex Networks

The availability of large-scale network data has given rise to the opportunity to investigate higher level organization of these networks using graph theoretic analysis. In this paper, we demonstrate a novel network decomposition tool called FacetsViewer in order to make sense of the deluge of network data. In contrast to traditional graph clustering techniques, it finds not just a single decomposition of the network, but a multi-faceted atlas of semantically meaningful decompositions that portray alternative perspectives of the landscape of the underlying network. Each facet in the atlas represents a distinct interpretation of how the network can be meaningfully decomposed and organized. To this end, FacetsViewer maximizes interpretative value of the atlas by optimizing inter-facet semantic and structural orthogonality. Specifically, we demonstrate various features of FacetsViewer and its superior ability to generate and visualize multi-faceted atlas of complex networks.