Elucidating tissue specific genes using the Benford distribution

Abstract

BackgroundThe RNA-seq technique is applied for the investigation of transcriptional behaviour. The reduction in sequencing costs has led to an unprecedented trove of gene expression data from diverse biological systems. Subsequently, principles from other disciplines such as the Benford law, which can be properly judged only in data-rich systems, can now be examined on this high-throughput transcriptomic information. The Benford law, states that in many count-rich datasets the distribution of the first significant digit is not uniform but rather logarithmic.ResultsAll tested digital gene expression datasets showed a Benford-like distribution when observing an entire gene set. This phenomenon was conserved in development and does not demonstrate tissue specificity. However, when obedience to the Benford law is calculated for individual expressed genes across thousands of cells, genes that best and least adhere to the Benford law are enriched with tissue specific or cell maintenance descriptors, respectively. Surprisingly, a positive correlation was found between the obedience a gene exhibits to the Benford law and its expression level, despite the former being calculated solely according to first digit frequency while totally ignoring the expression value itself. Nevertheless, genes with low expression that exhibit Benford behavior demonstrate tissue specific associations. These observations were extended to predict the likelihood of tissue specificity based on Benford behaviour in a supervised learning approach.ConclusionsThese results demonstrate the applicability and potential predictability of the Benford law for gleaning biological insight from simple count data.Electronic supplementary materialThe online version of this article (doi:10.1186/s12864-016-2921-x) contains supplementary material, which is available to authorized users.

Characterizing Human Cell Types and Tissue Origin Using the Benford Law

“…In all types of cells, the greatest overlap was between high-MAE genes and low mean-EXP genes, whereas all other groups showed almost no overlap (Figure 7). The relatively high overlap between high-MAE and low mean-EXP genes prompted us to explore the expression level distribution of the 200 highest and 200 lowest MAE genes (Figure 8); this analysis revealed that low-MAE genes have a wide expression distribution and are highly expressed, while high-MAE genes have a narrow expression distribution and are lowly expressed, as have been shown previously [13].…”

“…Since each tissue comprises many types of cells, we next tested the ability of our Benford algorithm to distinguish between different tissues and to detect the origin of a given tissue sample. We previously found, for 16 tissues that are included in the Illumina Human BodyMap 2.0 dataset (measured using microarray technology), that bulk gene expression data follow the Benford distribution when all genes are included in the calculation [13]. Therefore, we repeated this analysis, this time using GTEx bulk RNA data from 53 different tissues; similar to our previous results, we found that the first-digit distribution of the tissues adheres to the Benford distribution (Figure S8).…”

“…Previous studies have revealed that leading digits of gene expression data follow the Benford distribution—both for microarray intensities [7] and for RNA-seq digital expression levels—in a manner that is stable across tissues and is robust to different normalizations [13]. Deviations from the Benford distribution were detected when different gene sets with unique characteristics were explored, demonstrating that the likelihood of tissue specificity can be predicted based on the Benford behavior of the genes in question [13]. Motivated by this finding, in the present study, we tested whether the adherence of gene sets to the Benford distribution can serve as a means by which to classify cells to their cell type and to detect the origin of a tissue sample.…”

Broad distribution spectrum from Gaussian to power law appears in stochastic variations in RNA-seq data

“…An analysis of RNA-seq data from a two-condition, 48-replicate experiment using S. cerevisiae revealed that variations in expression levels for each gene conformed to both log-normal and NB distributions 36 . Beta-binomial and Benford distributions have been proposed for fitting gene expression data obtained by RNA-seq 37 , 38 . However, the physiological basis of these distributions and the significance of associated parameters remain unclear.…”

Gauss-power mixing distributions comprehensively describe stochastic variations in RNA-seq data

“…An analysis of RNA-seq data from a two-condition, 48-replicate experiment using S. cerevisiae revealed that variations in expression levels for each gene conformed to both log-normal and NB distributions (Gierlin’ski et al, 2015). Beta-binomial and Benford distributions have been proposed for fitting gene expression data obtained by RNA-seq (Smith et al, 2016; Karthik et al, 2016). However, the physiological basis of these distributions and the significance of associated parameters remain unclear.…”