Quality versus accuracy: result of a reanalysis of protein-binding microarrays from the DREAM5 challenge by using BayesPI2 including dinucleotide interdependence

BayesPI-BAR2: A New Python Package for Predicting Functional Non-coding Mutations in Cancer Patient Cohorts

Most of somatic mutations in cancer occur outside of gene coding regions. These mutations may disrupt the gene regulation by affecting protein-DNA interaction. A study of these disruptions is important in understanding tumorigenesis. However, current computational tools process DNA sequence variants individually, when predicting the effect on protein-DNA binding. Thus, it is a daunting task to identify functional regulatory disturbances among thousands of mutations in a patient. Previously, we have reported and validated a pipeline for identifying functional non-coding somatic mutations in cancer patient cohorts, by integrating diverse information such as gene expression, spatial distribution of the mutations, and a biophysical model for estimating protein binding affinity. Here, we present a new user-friendly Python package BayesPI-BAR2 based on the proposed pipeline for integrative whole-genome sequence analysis. This may be the first prediction package that considers information from both multiple mutations and multiple patients. It is evaluated in follicular lymphoma and skin cancer patients, by focusing on sequence variants in gene promoter regions. BayesPI-BAR2 is a useful tool for predicting functional non-coding mutations in whole genome sequencing data: it allows identification of novel transcription factors (TFs) whose binding is altered by non-coding mutations in cancer. BayesPI-BAR2 program can analyze multiple datasets of genome-wide mutations at once and generate concise, easily interpretable reports for potentially affected gene regulatory sites. The package is freely available at http://folk.uio.no/junbaiw/BayesPI-BAR2/.

Comparison of ChIP-Seq Data and a Reference Motif Set for Human KRAB C2H2 Zinc Finger Proteins

KRAB C2H2 zinc finger proteins (KZNFs) are the largest and most diverse family of human transcription factors, likely due to diversifying selection driven by novel endogenous retroelements (EREs), but the vast majority lack binding motifs or functional data. Two recent studies analyzed a majority of the human KZNFs using either ChIP-seq (60 proteins) or ChIP-exo (221 proteins) in the same cell type (HEK293). The ChIP-exo paper did not describe binding motifs, however. Thirty-nine proteins are represented in both studies, enabling the systematic comparison of the data sets presented here. Typically, only a minority of peaks overlap, but the two studies nonetheless display significant similarity in ERE binding for 32/39, and yield highly similar DNA binding motifs for 23 and related motifs for 34 (MoSBAT similarity score >0.5 and >0.2, respectively). Thus, there is overall (albeit imperfect) agreement between the two studies. For the 242 proteins represented in at least one study, we selected a highest-confidence motif for each protein, utilizing several motif-derivation approaches, and evaluating motifs within and across data sets. Peaks for the majority (158) are enriched (96% with AUC >0.6 predicting peak vs. nonpeak) for a motif that is supported by the C2H2 “recognition code,” consistent with intrinsic sequence specificity driving DNA binding in cells. An additional 63 yield motifs enriched in peaks, but not supported by the recognition code, which could reflect indirect binding. Altogether, these analyses validate both data sets, and provide a reference motif set with associated quality metrics.

Predicting Variation of DNA Shape Preferences in Protein-DNA Interaction in Cancer Cells with a New Biophysical Model

DNA shape readout is an important mechanism of transcription factor target site recognition, in addition to the sequence readout. Several machine learning-based models of transcription factor-DNA interactions, considering DNA shape features, have been developed in recent years. Here, we present a new biophysical model of protein-DNA interactions by integrating the DNA shape properties. It is based on the neighbor dinucleotide dependency model BayesPI2, where new parameters are restricted to a subspace spanned by the dinucleotide form of DNA shape features. This allows a biophysical interpretation of the new parameters as a position-dependent preference towards specific DNA shape features. Using the new model, we explore the variation of DNA shape preferences in several transcription factors across various cancer cell lines and cellular conditions. The results reveal that there are DNA shape variations at FOXA1 (Forkhead Box Protein A1) binding sites in steroid-treated MCF7 cells. The new biophysical model is useful for elucidating the finer details of transcription factor-DNA interaction, as well as for predicting cancer mutation effects in the future.

Integrative whole-genome sequence analysis reveals roles of regulatory mutations in BCL6 and BCL2 in follicular lymphoma

The contribution of mutations in regulatory regions to tumorigenesis has been the subject of many recent studies. We propose a new framework for integrative analysis of genome-wide sequencing data by considering diverse genetic information. This approach is applied to study follicular lymphoma (FL), a disease for which little is known about the contribution of regulatory gene mutations. Results from a test FL cohort revealed three novel highly recurrent regulatory mutation blocks near important genes implicated in FL, BCL6 and BCL2. Similar findings were detected in a validation FL cohort. We also found transcription factors (TF) whose binding may be disturbed by these mutations in FL: disruption of FOX TF family near the BCL6 promoter may result in reduced BCL6 expression, which then increases BCL2 expression over that caused by BCL2 gene translocation. Knockdown experiments of two TF hits (FOXD2 or FOXD3) were performed in human B lymphocytes verifying that they modulate BCL6/BCL2 according to the computationally predicted effects of the SNVs on TF binding. Overall, our proposed integrative analysis facilitates non-coding driver identification and the new findings may enhance the understanding of FL.

BayesPI-BAR: a new biophysical model for characterization of regulatory sequence variations

Sequence variations in regulatory DNA regions are known to cause functionally important consequences for gene expression. DNA sequence variations may have an essential role in determining phenotypes and may be linked to disease; however, their identification through analysis of massive genome-wide sequencing data is a great challenge. In this work, a new computational pipeline, a Bayesian method for protein–DNA interaction with binding affinity ranking (BayesPI-BAR), is proposed for quantifying the effect of sequence variations on protein binding. BayesPI-BAR uses biophysical modeling of protein–DNA interactions to predict single nucleotide polymorphisms (SNPs) that cause significant changes in the binding affinity of a regulatory region for transcription factors (TFs). The method includes two new parameters (TF chemical potentials or protein concentrations and direct TF binding targets) that are neglected by previous methods. The new method is verified on 67 known human regulatory SNPs, of which 47 (70%) have predicted true TFs ranked in the top 10. Importantly, the performance of BayesPI-BAR, which uses principal component analysis to integrate multiple predictions from various TF chemical potentials, is found to be better than that of existing programs, such as sTRAP and is-rSNP, when evaluated on the same SNPs. BayesPI-BAR is a publicly available tool and is able to carry out parallelized computation, which helps to investigate a large number of TFs or SNPs and to detect disease-associated regulatory sequence variations in the sea of genome-wide noncoding regions.

Comprehensive genome-wide transcription factor analysis reveals that a combination of high affinity and low affinity DNA binding is needed for human gene regulation

Background

High-throughput in vivo protein-DNA interaction experiments are currently widely used in gene regulation studies. Hitherto, comprehensive data analysis remains a challenge and for that reason most computational methods only consider the top few hundred or thousand strongest protein binding sites whereas weak protein binding sites are completely ignored.

Results

A new biophysical model of protein-DNA interactions, BayesPI2+, was developed to address the above-mentioned challenges. BayesPI2+ can be run in either a serial computation model or a parallel ensemble learning framework. BayesPI2+ allowed us to analyze all binding sites of the transcription factors, including weak binding that cannot be analyzed by other models. It is evaluated in both synthetic and real in vivo protein-DNA binding experiments. Analysing ESR1 and SPIB in breast carcinoma and activated B cell-like diffuse large B-cell lymphoma cell lines, respectively, revealed that the concerted binding to high and low affinity sites correlates best with gene expression.

Conclusions

BayesPI2+ allows us to analyze transcription factor binding on a larger scale than hitherto achieved. By this analysis, we were able to demonstrate that genes are regulated by concerted binding to high and low affinity binding sites. The program and output results are publicly available at: http://folk.uio.no/junbaiw/BayesPI2Plus.