A Comparative Study of Supervised Machine Learning Algorithms for the Prediction of Long-Range Chromatin Interactions

Abstract

The role of three-dimensional genome organization as a critical regulator of gene expression has become increasingly clear over the last decade. Most of our understanding of this association comes from the study of long range chromatin interaction maps provided by Chromatin Conformation Capture-based techniques, which have greatly improved in recent years. Since these procedures are experimentally laborious and expensive, in silico prediction has emerged as an alternative strategy to generate virtual maps in cell types and conditions for which experimental data of chromatin interactions is not available. Several methods have been based on predictive models trained on one-dimensional (1D) sequencing features, yielding promising results. However, different approaches vary both in the way they model chromatin interactions and in the machine learning-based strategy they rely on, making it challenging to carry out performance comparison of existing methods. In this study, we use publicly available 1D sequencing signals to model cohesin-mediated chromatin interactions in two human cell lines and evaluate the prediction performance of six popular machine learning algorithms: decision trees, random forests, gradient boosting, support vector machines, multi-layer perceptron and deep learning. Our approach accurately predicts long-range interactions and reveals that gradient boosting significantly outperforms the other five methods, yielding accuracies of about 95%. We show that chromatin features in close genomic proximity to the anchors cover most of the predictive information, as has been previously reported. Moreover, we demonstrate that gradient boosting models trained with different subsets of chromatin features, unlike the other methods tested, are able to produce accurate predictions. In this regard, and besides architectural proteins, transcription factors are shown to be highly informative. Our study provides a framework for the systematic prediction of long-range chromatin interactions, identifies gradient boosting as the best suited algorithm for this task and highlights cell-type specific binding of transcription factors at the anchors as important determinants of chromatin wiring mediated by cohesin.

Development of Single-Cell Transcriptomics and Its Application in COVID-19

“…Since then, single-cell sequencing technology has advanced significantly. Currently, single-cell research is mostly concerned with the transcriptome [ 9 ], spatial transcriptome [ 10 ], epigenetics [ 11 ] (DNA methylation [ 12 ], chromatin accessibility [ 13 ], chromatin interactions [ 14 ], histone modifications [ 15 ], and histone marks [ 16 ]), etc. The present article focuses mostly on single-cell transcriptomic research.…”

On the Unfounded Enthusiasm for Soft Selective Sweeps III: The Supervised Machine Learning Algorithm That Isn’t

“…The immense success that SML techniques have had in the world of e-commerce [ 10 ], as well as in some areas of medicine [ 11 , 12 ], genetics [ 13 ], genomics [ 14 ], and biochemistry [ 15 ] may have prompted Schrider and Kern to develop so-called supervised machine learning methodologies [ 3 ] to address evolutionary questions, particularly those aimed to clarify the relative importance of selection and random genetic drift during the evolution of genomes [ 4 ]. Such studies have been ballyhooed as “sophisticated”, “cutting-edge”, “robust”, and “valuable”, and it has been argued that they “make a strong case for the idea that machine learning methods could be useful for addressing diverse questions in molecular evolution” [ 16 ].…”

Computational methods for identifying enhancer‐promoter interactions

“…Numerous review articles have been published in recent decades concerning: enhancer interactions, including their role [21] at the genome-wide level; transcription enhancers in animal development, evolution [22], and disease [23]; functional contributions to transcription [24,25]; the functional significance of enhancer chromatin modification [26]; models that describe dynamic three-dimensional chromosome topology related to development enhancers; methods for identifying enhancer target genes [27] and enhancers [28][29][30]; the mechanisms of EPIs in higher eukaryotes [31]; bioinformatics analysis methods related to EPIs prediction [32][33][34][35]; analysis from sequence data [36,37]; and how EPIs control gene expression [38]. However, with the advancement of computational methods in the past decade, research has increasingly proposed methods for detecting enhancer-promoter interaction tools based on traditional machine learning or deep learning, but there has yet to be a global overview of solutions specifically for EPI identification.…”