Biology subjects

Georgakilas, G. K.

Publications and source records attributed to Georgakilas, G. K..

4 recordsLinked to original sources

bioRxiv2021

Bias in miRNA enrichment analysis related to gene functional annotations

BackgroundmiRNA functional enrichment is a type of analysis that is used to predict which biological functions may be affected by a group of miRNAs or validate whether a list of dysregulated miRNAs are linked to a diseased state. The standard method for functional enrichment analysis uses the hypergeometric distribution to produce p-values, depicting the strength of the association between a group of miRNAs and a biological function. However, in 2015, it was shown that this approach suffers from a bias related to miRNA targets produced by target prediction algorithms and a new randomization test was proposed to alleviate this issue. ResultsWe demonstrate the existence of another previously unreported underlying bias which affects gene annotation data sets; additionally, we show that the statistical measure used for the established randomization test is not sensitive enough to account for it. In this context, we show that the use of Jaccard coefficient (an alternative statistical measure) is able to alleviate the aforementioned issue. ConclusionsIn this paper, we illustrate the existence of a new bias affecting the miRNA functional enrichment analysis. This bias makes Fishers exact test unsuitable for miRNA functional enrichment analyses and there is also a need to adjust the established unbiased test accordingly. We propose the use of a modified version of the established test and in order to facilitate its use, we introduce a novel unbiased miRNA enrichment analysis tool that implements the proposed method. At the same time, by leveraging bit vectors, our tool guarantees fast and scalable execution. AvailabilityAll datasets used in the experiments throughout this paper are openly accessible on Zenodo (https://doi.org/10.5281/zenodo.5175819).

bioinformatics↗

bioRxiv2019

Solving the transcription start site identification problem with ADAPT-CAGE: a Machine Learning algorithm for analysis of CAGE data

Cap Analysis of Gene Expression (CAGE) experimental protocol has emerged as a powerful experimental technique for assisting in the identification of transcription start sites (TSSs). There is strong evidence that CAGE also identifies capping sites along various other locations of transcribed loci such as splicing byproducts, alternative isoforms and capped molecules overlapping introns and exons. We present ADAPT-CAGE, a Machine Learning framework which is trained to distinguish between CAGE signal derived from TSSs and transcriptional noise. ADAPT-CAGE provides annotation-agnostic, highly accurate and single-nucleotide resolution experimentally derived TSSs on a genome-wide scale. It has been specifically designed aiming for flexibility and ease-of-use by only requiring aligned CAGE data and the underlying genomic sequence. When compared to existing algorithms, ADAPT-CAGE exhibits improved performance on every benchmark that we designed based on both annotation- and experimentally-driven strategies. This performance boost brings ADAPT-CAGE in the spotlight as a computational framework that is able to assist in the refinement of gene regulatory networks, the incorporation of accurate information of gene expression regulators and alternative promoter usage in both physiological and pathological conditions.

bioinformatics↗

bioRxiv2019

MuStARD: Deep Learning for intra- and inter-species scanning of functional genomic patterns

Regions of the genome that produce different classes of functional elements also exhibit different patterns in their sequence, secondary structure, and evolutionary conservation. Deep Learning is a family of Machine Learning algorithms recently applied to a variety of pattern recognition problems. Here we present MuStARD (gitlab.com/RBP_Bioinformatics/mustard) a Deep Learning framework that can learn and combine sequence, structure, and conservation patterns in sets of functional regions, and accurately identify additional members of the given set over wide genomic areas. MuStARD is designed with general use in mind, and has sophisticated iterative fully-automated background selection capability. We demonstrate that MuStARD can be trained without changes on different classes of human small RNA loci (pre-microRNAs and snoRNAs) and accurately build prediction models for both, outperforming state of the art methods specifically designed for each specific class. Furthermore, we demonstrate the ability of MuStARD for inter-species identification of functional elements by predicting mouse small RNAs using human trained models. MuStARD is easy to deploy and extend to a variety of genomic classification questions.

bioinformatics↗

bioRxiv2018

A cosine similarity-based method to infer variability of chromatin accessibility at the single-cell level

Cellular identity between generations of developing cells is propagated through the epigenome particularly via the accessible parts of the chromatin. It is now possible to measure chromatin accessibility at single-cell resolution using single-cell assay for transposase accessible chromatin (scATAC-seq), which can reveal the regulatory variation behind the phenotypic variation. However, single-cell chromatin accessibility data are sparse, binary, and high dimensional, leading to unique computational challenges. To overcome these difficulties, we developed PRISM a computational workflow and R package (https://github.com/stanleycai123/PRISM) that quantifies cell-to-cell chromatin accessibility variation while controlling for technical biases. Using data generated in our lab or publically available, we show that PRISM outperforms an existing algorithm, which relies on the aggregate of signal across a set of genomic regions. PRISM shows robustness to noise in low accessibility cells and reveals previously masked accessibility variation where accessible sites differ between cells but total number of accessible sites is constant. We also show that PRISM, but not an existing algorithm, finds suppressed heterogeneity of accessibility at CTCF binding sites. PRISM is a novel multidimensional scaling-based method using angular cosine distance metrics coupled with distance from the spatial centroid. PRISM takes differences in accessibility at each genomic region between single cells into account. This updated approach uncovers new biological results with profound implications on the cellular heterogeneity of chromatin architecture.

genomics↗