Dataset Information

High dimensional model representation of log-likelihood ratio: binary classification with expression data.

ABSTRACT:

Background

Binary classification rules based on a small-sample of high-dimensional data (for instance, gene expression data) are ubiquitous in modern bioinformatics. Constructing such classifiers is challenging due to (a) the complex nature of underlying biological traits, such as gene interactions, and (b) the need for highly interpretable glass-box models. We use the theory of high dimensional model representation (HDMR) to build interpretable low dimensional approximations of the log-likelihood ratio accounting for the effects of each individual gene as well as gene-gene interactions. We propose two algorithms approximating the second order HDMR expansion, and a hypothesis test based on the HDMR formulation to identify significantly dysregulated pairwise interactions. The theory is seen as flexible and requiring only a mild set of assumptions.

Results

We apply our approach to gene expression data from both synthetic and real (breast and lung cancer) datasets comparing it also against several popular state-of-the-art methods. The analyses suggest the proposed algorithms can be used to obtain interpretable prediction rules with high prediction accuracies and to successfully extract significantly dysregulated gene-gene interactions from the data. They also compare favorably against their competitors across multiple synthetic data scenarios.

Conclusion

The proposed HDMR-based approach appears to produce a reliable classifier that additionally allows one to describe how individual genes or gene-gene interactions affect classification decisions. Both real and synthetic data analyses suggest that our methods can be used to identify gene networks with dysregulated pairwise interactions, and are therefore appropriate for differential networks analysis.

SUBMITTER: Foroughi Pour A

PROVIDER: S-EPMC7183128 | biostudies-literature | 2020 Apr

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

High dimensional model representation of log-likelihood ratio: binary classification with expression data.

Foroughi Pour Ali A Pietrzak Maciej M Dalton Lori A LA Rempała Grzegorz A GA

BMC bioinformatics 20200425 1

<h4>Background</h4>Binary classification rules based on a small-sample of high-dimensional data (for instance, gene expression data) are ubiquitous in modern bioinformatics. Constructing such classifiers is challenging due to (a) the complex nature of underlying biological traits, such as gene interactions, and (b) the need for highly interpretable glass-box models. We use the theory of high dimensional model representation (HDMR) to build interpretable low dimensional approximations of the log- ...[more]

PMID: 32334509

Similar Datasets

Project description:BackgroundDeveloping binary classification rules based on SNP observations has been a major challenge for many modern bioinformatics applications, e.g., predicting risk of future disease events in complex conditions such as cancer. Small-sample, high-dimensional nature of SNP data, weak effect of each SNP on the outcome, and highly non-linear SNP interactions are several key factors complicating the analysis. Additionally, SNPs take a finite number of values which may be best understood as ordinal or categorical variables, but are treated as continuous ones by many algorithms.MethodsWe use the theory of high dimensional model representation (HDMR) to build appropriate low dimensional glass-box models, allowing us to account for the effects of feature interactions. We compute the second order HDMR expansion of the log-likelihood ratio to account for the effects of single SNPs and their pairwise interactions. We propose a regression based approach, called linear approximation for block second order HDMR expansion of categorical observations (LABS-HDMR-CO), to approximate the HDMR coefficients. We show how HDMR can be used to detect pairwise SNP interactions, and propose the fixed pattern test (FPT) to identify statistically significant pairwise interactions.ResultsWe apply LABS-HDMR-CO and FPT to synthetically generated HAPGEN2 data as well as to two GWAS cancer datasets. In these examples LABS-HDMR-CO enjoys superior accuracy compared with several algorithms used for SNP classification, while also taking pairwise interactions into account. FPT declares very few significant interactions in the small sample GWAS datasets when bounding false discovery rate (FDR) by 5%, due to the large number of tests performed. On the other hand, LABS-HDMR-CO utilizes a large number of SNP pairs to improve its prediction accuracy. In the larger HAPGEN2 dataset FTP declares a larger portion of SNP pairs used by LABS-HDMR-CO as significant.ConclusionLABS-HDMR-CO and FPT are interesting methods to design prediction rules and detect pairwise feature interactions for SNP data. Reliably detecting pairwise SNP interactions and taking advantage of potential interactions to improve prediction accuracy are two different objectives addressed by these methods. While the large number of potential SNP interactions may result in low power of detection, potentially interacting SNP pairs, of which many might be false alarms, can still be used to improve prediction accuracy.

Project description:BACKGROUND:In genomics and proteomics, membrane protein analysis have shown that such analyses are very important to support the understanding of complex biological processes. In Genome-wide investigations of membrane proteins a large number of short, distinct sequence motifs has been revealed. Such motifs found so far support the understanding of the folded membrane protein in the membrane environment. They provide important information about functional or stabilizing properties. Recently several integrative approaches have been proposed to extract meaningful information out of the membrane environment. However, many information based approaches deliver results having deficits of visualisation outputs. Outgoing from high-throughput protein data analysis, these outputs play an important role in the evaluation of high-dimensional protein data, to establish a biological relationship and ultimately to provide useful information for research. RESULTS:We have evaluated different resulting graphs generated from statistical analysis of consecutive motifs in helical structures of the membrane environment. Our results show that representative motifs with high occurrence in all investigated protein families are responsible for the general importance in alpha-helical membrane structure formation. Further, motifs which often occur with others in their function as so called "hubs" lead to the assumption, that these motifs constitute as important components in helical structures within the membrane. Otherwise, consecutive motifs and hubs which show a high occurrence in certain families only can be classified as important for family-specific functional characteristics. Summarized, we are able to bridge our graphical results from high-throughput analysis of membrane proteins over networking with databases to a biological context. CONCLUSIONS:Our results and the corresponding graphical visualisation support the understanding and interpretation of structure forming and functional motifs of membrane proteins. Our results are useful to interpret and refine results of common developed approaches. At last we show a simple way to visualise high-dimensional protein data in context to biological relevant information.

Project description:During commitment of a multipotent stem or progenitor cell to a particular lineage, a large number of genes alter their expression in a coordinated manner orchestrated by the gene regulatory network (GRN). The constraints imposed by the GRN govern how cells move in the high-dimensional gene expression state space and can be understood as a dynamical system in which phenotypic cell states (cell types) are attractors that stabilize the cell-type characteristic gene expression pattern against molecular noise. Despite insights from various theoretical models, it remains elusive how multipotent cells, when committing to a specific lineage, exit their attractor and enter a new distinct attractor. Here we show, using single-cell resolution monitoring of transcript patterns by qPCR that commitment of multipotent blood progenitor cells to either the erythroid or the myeloid lineage is preceded by a destabilization of the progenitorsâ?? attractor state and a slowing-down of relaxation of cells from outlier states, indicating a critical state transition (â??tipping pointâ??). The high-dimensionality of the system (many genes) and availability of individual trajectories of a large ensemble of systems (many cells) affords a novel signature for critical transition which can be predicted from theory: Decrease of correlation between cells and concomitant increase of correlation between genes as the cell population approaches the tipping point. Consistent with a destabilizing bifurcation that simultaneously opens access to the erythroid and myeloid attractors, differentiation signal for either lineage caused some cells to commit to the â??wrongâ?? fate; moreover providing conflicting signals resulted in a delayed decision at the bifurcation point that however was ultimately resolved by commitment to one fate. These results suggest that the theoretical framework of â??early-warning signsâ?? and critical transitions can be applied to ensembles of high-dimensional systems, offering a formal tool for analyzing single-cell omics data beyond current descriptive computational pattern recognition. Mouse blood progenitor cells (EML cell line) was exposed to EPO, IL-3/GM-CSF or a mixture of both cytokines and gene expression change was measured in sorted subpopulations wrt Sca1 progenitor surface marker expression. In total, there was 4 conditions (including control), three time points (including d0) and 20 samples (10 samples in duplicates) were analyzed. Two independent experiments were performed for each condition. The untreated progenitor cell population was used as control.

Dataset Information

High dimensional model representation of log-likelihood ratio: binary classification with expression data.

Background

Results

Conclusion

Publications

High dimensional model representation of log-likelihood ratio: binary classification with expression data.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets