Michael Sharpnack

Visualizing Multidimensional Data with Glyph SPLOMs

Scatterplot matrices or SPLOMs provide a feasible method of visualizing and representing multi‐dimensional data especially for a small number of dimensions. For very high dimensional data, we introduce a novel technique to summarize a SPLOM, as a clustered matrix of glyphs, or a Glyph SPLOM. Each glyph visually encodes a general measure of dependency strength, distance correlation, and a logical dependency class based on the occupancy of the scatterplot quadrants. We present the Glyph SPLOM as a general alternative to the traditional correlation based heatmap and the scatterplot matrix in two examples: demography data from the World Health Organization (WHO), and gene expression data from developmental biology. By using both, dependency class and strength, the Glyph SPLOM illustrates high dimensional data in more detail than a heatmap but with more summarization than a SPLOM. More importantly, the summarization capabilities of Glyph SPLOM allow for the assertion of “necessity” causal relationships in the data and the reconstruction of interaction networks in various dynamic systems.

DISCOVERY OF MOLECULARLY TARGETED THERAPIES

The delivery of personalized healthcare is predicated on the application of the best available scientific knowledge to the practice of medicine in order to promote health, improve outcomes and enhance patient safety [1-3]. Unfortunately, current approaches to basic science research and clinical care are poorly integrated, yielding clinical decision-making processes that do not take advantage of up-to-date scientific knowledge [2-4]. Basic scientists investigating the biological basis for a given disease may regularly encounter synergistic effects spanning two or more bio-molecular entities or processes that can contribute to our understanding of the mechanisms underlying phenomena such as the etiologic basis of the targeted disease state or potential response to therapeutic agents [5]. However, systematic approaches to the use of that knowledge in order to directly inform the selection of targeted molecular therapies for “real world” patients are extremely limited [1, 3, 6-9]. There are an increasing number of multi-modelling and in-silico knowledge synthesis techniques that can provide investigators with the tools to quickly generate hypotheses concerning the relationships between entities found in heterogeneous collections of scientific data — for example, exploring potential linkages among genes, phenotypes and molecularly targeted therapeutic agents, thus enabling the “forward engineering” of treatment strategies based on knowledge generated via basic science studies [1, 4, 6, 10, 11]. Ultimately, the goal of such methodologies is to accelerate the identification of actionable research questions that can make direct contributions to clinical practice. Given increasing concerns over the barriers to the timely translation of discoveries from the laboratory to the clinic or broader population settings, such high-throughput hypothesis generation and testing is highly desirable [1, 4, 6, 8, 12]. These needs are particularly critical in numerous disease areas where the availability of new therapeutic agents is constrained, thus calling for the re-use and repositioning of existing treatments [13, 14]. In response to the challenges and opportunities enumerated above, there exits an emerging body of research and development focusing on multi-modeling approaches to the discovery of molecularly targeted therapies, including experimental paradigms spanning a spectrum from the identification of molecular targets for drugs, to the repurposing or repositioning of existing agents that utilize such targets, to the systematic identification of novel combination therapy regimens that amplify or enhance the effectiveness of their constituent components. This focus is motivated by recent and significant advances in the state of systems biology and medicine that have demonstrated that the ability to generate and reason across complex and scalar models is essential to the discovery of high-impact biologically and clinically actionable knowledge [1, 4, 12]. Such approaches are designed to overcome the limitations of reductionist approaches to scientific discovery, replacing decomposition-focused problem-solving with integrative network-based modeling and analysis techniques [4, 8]. Systems-level analysis of complex problem domains ultimately enables the study of critical interactions that influence health and wellness across a scale from molecules to populations, and are not observable when such systems are broken down into constituent components. The use of systems-level analysis methodologies is well supported by the foundational theory of vertical reasoning first proposed by Blois [15]. This theory holds that effective decision-making in the biomedical domain is predicated on the vertical integration of multiple, scalar levels of reasoning. This fundamental premise is the basis for a correlative framework put forth by Tsafnat and colleagues, which states that the ability to replicate expert reasoning relative to complex biomedical problems using computational agents (e.g., in-silico knowledge synthesis) requires the replication of such multi-scalar and integrative decision-making [16]. In order to achieve such an outcome, Tsafnat posits that multi-scalar decision-making in an in-silico context requires both: 1) the generation of component decision-making models at multiple scales; and 2) the similar generation of interchange layers that define important pair-wise connections between entities situated in two or more component models, often referred to as vertical linkages [16]. When such component models and interchange layers are combined in a computationally actionable format, they yield what can be referred to as a multi-model for a given domain that is able to satisfy the premises of Blois’ vertical reasoning axiom, and therefore facilitate the replication of expert performance in a high-throughput manner [16]. Of note, this type of approach is extremely reliant upon graph-theoretic reasoning and representational models, using a network paradigm that allows for the application of logical reasoning operations spanning the entities and relationships that make up a multi-model [8]. Network paradigms have been regularly shown to be the ideal representational model for naturally occurring systems, such as the ‘scale-free’ networks encountered in biological and clinical phenomena [8]. At the most basic level, network-based multi-modeling across scales presents an elegant and computationally tractable approach to understanding and evaluating complex biological and clinical systems in order to discover the knowledge incumbent to such constructs. This type of approach benefits from a robust set of foundational theories and frameworks that can inform and shape the application of multi-modeling techniques to a variety of knowledge discovery use cases. As such, there is a growing body of evidence concerning the application of network-based approaches to multi-modeling with an emphasis on therapeutic agent discovery, re-positioning and molecular targeting. Examples of such evidence include reports and perspectives published by Hood and Perlmutter [1], Butcher and colleagues [12], and Lussier and Chen [13].

Proteogenomic Analysis of Surgically Resected Lung Adenocarcinoma

Introduction: Despite apparently complete surgical resection, approximately half of resected early‐stage lung cancer patients relapse and die of their disease. Adjuvant chemotherapy reduces this risk by only 5% to 8%. Thus, there is a need for better identifying who benefits from adjuvant therapy, the drivers of relapse, and novel targets in this setting. Methods: RNA sequencing and liquid chromatography/liquid chromatography–mass spectrometry proteomics data were generated from 51 surgically resected non–small cell lung tumors with known recurrence status. Results: We present a rationale and framework for the incorporation of high‐content RNA and protein measurements into integrative biomarkers and show the potential of this approach for predicting risk of recurrence in a group of lung adenocarcinomas. In addition, we characterize the relationship between mRNA and protein measurements in lung adenocarcinoma and show that it is outcome specific. Conclusions: Our results suggest that mRNA and protein data possess independent biological and clinical importance, which can be leveraged to create higher‐powered expression biomarkers.

Integrative cancer patient stratification via subspace merging

MOTIVATION Technologies that generate high-throughput omics data are flourishing, creating enormous, publicly available repositories of multi-omics data. As many data repositories continue to grow, there is an urgent need for computational methods that can leverage these data to create comprehensive clusters of patients with a given disease. RESULTS Our proposed approach creates a patient-to-patient similarity graph for each data type as an intermediate representation of each omics data type and merges the graphs through subspace analysis on a Grassmann manifold. We hypothesize that this approach generates more informative clusters by preserving the complementary information from each level of omics data. We applied our approach to The Cancer Genome Atlas (TCGA) breast cancer dataset and show that by integrating gene expression, microRNA and DNA methylation data, our proposed method can produce clinically useful subtypes of breast cancer. We then investigate the molecular characteristics underlying these subtypes. We discover a highly expressed cluster of genes on chromosome 19p13 that strongly correlates with survival in TCGA breast cancer patients and validate these results in three additional breast cancer datasets. We also compare our approach with previous integrative clustering approaches and obtain comparable or superior results. AVAILABILITY AND IMPLEMENTATION https://github.com/michaelsharpnack/GrassmannCluster. SUPPLEMENTARY INFORMATION Supplementary data are available at Bioinformatics online.