A novel deep mining model for effective knowledge discovery from omics data

Abstract

Knowledge discovery from omics data has become a common goal of current approaches to personalised cancer medicine and understanding cancer genotype and phenotype. However, high-throughput biomedical datasets are characterised by high dimensionality and relatively small sample sizes with small signal-to-noise ratios. Extracting and interpreting relevant knowledge from such complex datasets therefore remains a significant challenge for the fields of machine learning and data mining. In this paper, we exploit recent advances in deep learning to mitigate against these limitations on the basis of automatically capturing enough of the meaningful abstractions latent with the available biological samples. Our deep feature learning model is proposed based on a set of non-linear sparse Auto-Encoders that are deliberately constructed in an under-complete manner to detect a small proportion of molecules that can recover a large proportion of variations underlying the data. However, since multiple projections are applied to the input signals, it is hard to interpret which phenotypes were responsible for deriving such predictions. Therefore, we also introduce a novel weight interpretation technique that helps to deconstruct the internal state of such deep learning models to reveal key determinants underlying its latent representations. The outcomes of our experiment provide strong evidence that the proposed deep mining model is able to discover robust biomarkers that are positively and negatively associated with cancers of interest. Since our deep mining model is problem-independent and data-driven, it provides further potential for this research to extend beyond its cognate disciplines.

mRNA Biomarkers for Invasive Breast Cancer based on a Deep Feature Selection Approach

“…Our fi ndings reveal strong evidence of a positive or a negative association between the discovered mRNA markers and oestrogen and progesterone receptors. In this work we applied a promising new deep feature selection approach, originally introduced by Alzubaidi, et al [19], to three separate breast invasive carcinoma datasets from the TCGA database with the aim of modelling and analysing gene expression data to discover interesting complex patterns that may appear to aid the development of new and innovative diagnostic and prognostic tools for ER+/PR+ invasive breast cancer. Given this stated aim, we conclude the research a success in that our model, with its deep feature extraction and feature selection modules, not only discovered sets of genes previously known to be associated with breast cancer but also a small panel of genes that appear to have largely gone unnoticed in the literature.…”

“…In this paper, we employ a promising new deep knowledge discovery model, defi ned by Alzubaidi, et al [19], with two fundamental components: i) a type of deep net referred to as a Stacked Sparse Compressed Auto-Encoder (SSCAE) that generates hierarchical abstract representations of the raw mRNA expression data; and ii) an eff ective weight interpretation method that determines the salient input genes that underlie the abstract and compressed representations formed by SSCAE. To avoid over fi tting, SSCAE uses a regularization term constraining the formation of its hidden state to promote under-complete representations during learning.…”

Medical analytics for healthcare intelligence – Recent advances and future directions

“…Knowledge discovery from omics data has become a common goal of current approaches to personalised cancer medicine and understanding cancer genotype and phenotype. With omics data characterised by high dimensionality and relatively small sample sizes with small signal-to-noise ratios, Alzubaidi et al [6] propose a deep feature learning model based on a set of non-linear sparse Auto-Encoders that are deliberately constructed in an under-complete manner to detect a small proportion of molecules that can recover a large proportion of variations underlying the data This is followed by the introduction of a novel weight interpretation technique that helps to deconstruct the internal state of such deep learning models to reveal key determinants underlying its latent rep-resentations. Experiments reveal that the discovered biomarkers demonstrate computational and biological relevance, as well as the capability to construct highly accurate and reliable prediction models.…”

Iterative Decorrelation Analysis, Unit of Measure Preserving Transformations and Latent Biomarker Discovery

“…Both unsupervised latent variable and supervised minimum redundancy approaches have been applied in biomarker discovery, with supervised methods currently dominating the field. Effective supervised alternatives for biomarker discovery include machine learning techniques, which can handle hyperdimensional data sets and mitigate false discovery rates by relying on stochastic approaches (9,24,25,26,27,28,29). These ML approaches may also identify biomarker signatures (latent variables) associated with disease progression, diagnosis, or treatment effects (30,31,32).…”