Flavia Moser

Inferring cancer subnetwork markers using density-constrained biclustering

Motivation: Recent genomic studies have confirmed that cancer is of utmost phenotypical complexity, varying greatly in terms of subtypes and evolutionary stages. When classifying cancer tissue samples, subnetwork marker approaches have proven to be superior over single gene marker approaches, most importantly in cross-platform evaluation schemes. However, prior subnetwork-based approaches do not explicitly address the great phenotypical complexity of cancer. Results: We explicitly address this and employ density-constrained biclustering to compute subnetwork markers, which reflect pathways being dysregulated in many, but not necessarily all samples under consideration. In breast cancer we achieve substantial improvements over all cross-platform applicable approaches when predicting TP53 mutation status in a well-established non-cross-platform setting. In colon cancer, we raise prediction accuracy in the most difficult instances from 87% to 93% for cancer versus non−cancer and from 83% to (astonishing) 92%, for with versus without liver metastasis, in well-established cross-platform evaluation schemes. Availability: Software is available on request. Contact: alexsch@math.berkeley.edu; ester@cs.sfu.ca Supplementary information: Supplementary data are available at Bioinformatics online.

Joint cluster analysis of attribute and relationship data withouta-priori specification of the number of clusters

In many applications, attribute and relationship data areavailable, carrying complementary information about real world entities. In such cases, a joint analysis of both types of data can yield more accurate results than classical clustering algorithms that either use only attribute data or only relationship (graph) data. The Connected k-Center (CkC) has been proposed as the first joint cluster analysis model to discover k clusters which are cohesive on both attribute and relationship data. However, it is well-known that prior knowledge on the number of clusters is often unavailable in applications such as community dentification and hotspot analysis. In this paper, we introduce and formalize the problem of discovering an a-priori unspecified number of clusters in the context of joint cluster analysis of attribute and relationship data, called Connected X Clusters (CXC) problem. True clusters are assumed to be compact and distinctive from their neighboring clusters in terms of attribute data and internally connected in terms of relationship data. Different from classical attribute-based clustering methods, the neighborhood of clusters is not defined in terms of attribute data but in terms of relationship data. To efficiently solve the CXC problem, we present JointClust, an algorithm which adopts a dynamic two-phase approach. In the first phase, we find so called cluster atoms. We provide a probability analysis for thisphase, which gives us a probabilistic guarantee, that each true cluster is represented by at least one of the initial cluster atoms. In the second phase, these cluster atoms are merged in a bottom-up manner resulting in a dendrogram. The final clustering is determined by our objective function. Our experimental evaluation on several real datasets demonstrates that JointClust indeed discovers meaningful and accurate clusterings without requiring the user to specify the number of clusters.

A method for multi-relational classification using single and multi-feature aggregation functions

This paper presents a novel method for multi-relational classification via an aggregation-based Inductive Logic Programming (ILP) approach. We extend the classical ILP representation by aggregation of multiple-features which aid the classification process by allowing for the analysis of relationships and dependencies between different features. In order to efficiently learn rules of this rich format, we present a novel algorithm capable of performing aggregation with the use of virtual joins of the data. By using more expressive aggregation predicates than the existential quantifier used in standard ILP methods, we improve the accuracy of multi-relational classification. This claim is supported by experimental evaluation on three different real world datasets.

Module discovery by exhaustive search for densely connected, co-expressed regions in biomolecular interaction networks.

Background Computational prediction of functionally related groups of genes (functional modules) from large-scale data is an important issue in computational biology. Gene expression experiments and interaction networks are well studied large-scale data sources, available for many not yet exhaustively annotated organisms. It has been well established, when analyzing these two data sources jointly, modules are often reflected by highly interconnected (dense) regions in the interaction networks whose participating genes are co-expressed. However, the tractability of the problem had remained unclear and methods by which to exhaustively search for such constellations had not been presented. Methodology/Principal Findings We provide an algorithmic framework, referred to as Densely Connected Biclustering (DECOB), by which the aforementioned search problem becomes tractable. To benchmark the predictive power inherent to the approach, we computed all co-expressed, dense regions in physical protein and genetic interaction networks from human and yeast. An automatized filtering procedure reduces our output which results in smaller collections of modules, comparable to state-of-the-art approaches. Our results performed favorably in a fair benchmarking competition which adheres to standard criteria. We demonstrate the usefulness of an exhaustive module search, by using the unreduced output to more quickly perform GO term related function prediction tasks. We point out the advantages of our exhaustive output by predicting functional relationships using two examples. Conclusion/Significance We demonstrate that the computation of all densely connected and co-expressed regions in interaction networks is an approach to module discovery of considerable value. Beyond confirming the well settled hypothesis that such co-expressed, densely connected interaction network regions reflect functional modules, we open up novel computational ways to comprehensively analyze the modular organization of an organism based on prevalent and largely available large-scale datasets. Availability Software and data sets are available at http://www.sfu.ca/~ester/software/DECOB.zip.

H3N2v and Other Influenza Epidemic Risk Based on Age-Specific Estimates of Sero-Protection and Contact Network Interactions

Cases of a novel swine-origin influenza A(H3N2) variant (H3N2v) have recently been identified in the US, primarily among children. We estimated potential epidemic attack rates (ARs) based on age-specific estimates of sero-susceptibility and social interactions. A contact network model previously established for the Greater Vancouver Area (GVA), Canada was used to estimate average epidemic (infection) ARs for the emerging H3N2v and comparator viruses (H1N1pdm09 and an extinguished H3N2 seasonal strain) based on typical influenza characteristics, basic reproduction number (R0), and effective contacts taking into account age-specific sero-protection rates (SPRs). SPRs were assessed in sera collected from the GVA in 2009 or earlier (pre-H1N1pdm09) and fall 2010 (post-H1N1pdm09, seasonal A/Brisbane/10/2007(H3N2), and H3N2v) by hemagglutination inhibition (HI) assay. SPR was assigned per convention based on proportion with HI antibody titre ≥40 (SPR40). Recognizing that the HI titre ≥40 was established as the 50%sero-protective threshold we also explored for ½SPR40, SPR80 and a blended gradient defined as: ¼SPR20, ½SPR40, ¾SPR80, SPR160. Base case analysis assumed R0 = 1.40, but we also explored R0 as high as 1.80. With R0 = 1.40 and SPR40, simulated ARs were well aligned with field observations for H1N1pdm09 incidence (AR: 32%), sporadic detections without a third epidemic wave post-H1N1pdm09 (negligible AR<0.1%) as well as A/Brisbane/10/2007(H3N2) seasonal strain extinction and antigenic drift replacement (negligible AR<0.1%). Simulated AR for the novel swine-origin H3N2v was 6%, highest in children 6–11years (16%). However, with modification to SPR thresholds per above, H3N2v AR ≥20% became possible. At SPR40, H3N2v AR ≥10%, ≥15% or ≥30%, occur if R0≥1.48, ≥1.56 or ≥1.86, respectively. Based on conventional assumptions, the novel swine-origin H3N2v does not currently pose a substantial pandemic threat. If H3N2v epidemics do occur, overall community ARs are unlikely to exceed typical seasonal influenza experience. However risk assessment may change with time and depends crucially upon the validation of epidemiological features of influenza, notably the serologic correlate of protection and R0.

Epidemic progression on networks based on disease generation time

We investigate the time evolution of disease spread on a network and present an analytical framework using the concept of disease generation time. Assuming a susceptible–infected–recovered epidemic process, this network-based framework enables us to calculate in detail the number of links (edges) within the network that are capable of producing new infectious nodes (individuals), the number of links that are not transmitting the infection further (non-transmitting links), as well as the number of contacts that individuals have with their neighbours (also known as degree distribution) within each epidemiological class, for each generation period. Using several examples, we demonstrate very good agreement between our analytical calculations and the results of computer simulations.