BoostMe accurately predicts DNA methylation values in whole-genome bisulfite sequencing of multiple human tissues

Zou, Luli S.; Erdos, Michael R.; Taylor, D. Leland; Chines, Peter S.; Varshney, Arushi; Parker, Stephen C. J.; Collins, Francis S.; Didion, John P.

doi:10.1101/207506

Cited by 6 publications

(4 citation statements)

References 124 publications

(232 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Imputation can be used to estimate the methylation value for each missing CpG site 15, 16 and thereby, can be used to help increase platform intra- and interoperability. We used BoostMe 16 to impute the vast majority (i.e. 99.95%) of missing CpG sites, leveraging information learnt from just the neighboring CpG within the same dataset.…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Performance comparison and in-silico harmonisation of commercial platforms for DNA methylome analysis by targeted bisulfite sequencing

Tanić

Moghul

Rodney

et al. 2021

Preprint

View full text Add to dashboard Cite

DNA methylation is a key epigenetic modification in the regulation of cell fate and differentiation, and its analysis is gaining increasing importance in both basic and clinical research. Targeted Bisulfite Sequencing (TBS) has become the method of choice for the cost-effective, targeted analysis of the human methylome at base-pair resolution. Here we benchmarked five commercially available TBS platforms, including three hybridization capture-based (Agilent, Roche, and Illumina) and two RRBS-based (Diagenode and NuGen), across 16 samples. A subset of these were also compared to whole-genome DNA methylation sequencing with the Illumina and Oxford Nanopore platforms. We assessed performance with respect to workflow complexity, on/off-target performance, coverage, accuracy and reproducibility. We find all platforms able to produce usable data but major differences for some performance criteria, especially in the number and identity of the CpG sites covered, which affects the interoperability of datasets generated on these different platforms. To overcome this limitation, we used imputation and show that it improves the interoperability from an average of 10.35% (0.8M CpG sites) to 97% (7.6M CpG sites). Our study provides cross-validated guidance on which TBS platform to use for different features of the methylome and offers an imputation-based harmonization solution for improved interoperability between platforms, allowing comparative and integrative analysis.

show abstract

Section: Resultsmentioning

confidence: 99%

“…We used BoostMe 16 to impute the vast majority (i.e. 99.95%) of missing CpG sites, leveraging information learnt from just the neighboring CpG within the same dataset.…”

Section: Methodsmentioning

confidence: 99%

Performance comparison and in-silico harmonisation of commercial platforms for DNA methylome analysis by targeted bisulfite sequencing

Tanić

Moghul

Rodney

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…For example, variational autoencoders have been successfully applied for dimensionality reduction of methylation data [8]. Other methods focus on imputation of single CpG sites in tissue samples using, among others, linear regression [9], Random Forests [10], autoencoders [11], gradient boosting [12] or mixture models [13]. In addition to using intrasample dependencies between neighboring CpG sites, some of these methods adopt the idea of leveraging information from multiple (tissue) samples for prediction [12, 13].…”

Section: Introductionmentioning

confidence: 99%

CpG Transformer for imputation of single-cell methylomes

Waele

Clauwaert

Menschaert

et al. 2021

Preprint

View full text Add to dashboard Cite

Motivation: The adoption of current single-cell DNA methylation sequencing protocols is hindered by incomplete coverage, outlining the need for effective imputation techniques. The task of imputing single-cell (methylation) data requires models to build an understanding of underlying biological processes. Current approaches compress intercellular methylation dependencies in some way and, hence, do not provide a general-purpose way of learning interactions between neighboring CpG sites both within- and between cells. Results: We adapt the transformer neural network architecture to operate on methylation matrices through the introduction of a novel 2D sliding window self-attention. The obtained CpG Transformer displays state-of-the-art performances on a wide range of scBS-seq and scRRBS-seq datasets. Furthermore, we demonstrate the interpretability of CpG Transformer and illustrate its rapid transfer learning properties, allowing practitioners to train models on new datasets with a limited computational and time budget. Availability and Implementation: CpG Transformer is freely available at https://github.com/gdewael/cpg-transformer.

show abstract

“…The technique was found to outperform other machine learning and deep learning techniques in many competitions such as Kaggle and KDDCup (Chen and Guestrin, 2016;Dhaliwal, et al, 2018), especially for datasets with sparse matrix. It has been successfully applied in many bioinformatic studies, such as miRNA-disease association (Chen, et al, 2018), protein translocation (Mendik, et al, 2019), protein-protein interactions (Basit, et al, 2018), and DNA methylation (Zou, et al, 2018).…”

Section: Introductionmentioning

confidence: 99%

Accurate prediction of genome-wide RNA secondary structure profile based on extreme gradient boosting

Ke¹,

Rao²,

Zhao

et al. 2020

Bioinformatics

View full text Add to dashboard Cite

Motivation RNA secondary structure plays a vital role in fundamental cellular processes, and identification of RNA secondary structure is a key step to understand RNA functions. Recently, a few experimental methods were developed to profile genome-wide RNA secondary structure, i.e. the pairing probability of each nucleotide, through high-throughput sequencing techniques. However, these high-throughput methods have low precision and can’t cover all nucleotides due to limited sequencing coverage. Results Here we have developed a new method for the prediction of genome-wide RNA secondary structure profile from RNA sequence based on the extreme Gradient Boosting technique. The method achieves predictions with areas under the receiver operating characteristic curve (AUC) greater than 0.9 on three different datasets, and AUC of 0.888 by an independent test on the recently released Zika virus data. These AUCs are consistently >5 % greater than the ones by the CROSS method recently developed based on a shallow neural network. Further analysis on the 1000 Genome Project data showed that our predicted unpaired probabilities are highly correlated (>0.8) with the minor allele frequencies at synonymous, non-synonymous mutations, and mutations in untranslated region, which were higher than those generated by RNAplfold. Moreover, the prediction over all human mRNA indicated a consistent result with previous observation that there is a periodic distribution of unpaired probability on codons. The accurate prediction by our method indicates that such model trained on genome-wide experimental data might be an alternative for analytical methods. Availability The GRASP is available for academic use at https://github.com/sysu-yanglab/GRASP. Supplementary information Supplementary data are available online.

show abstract

BoostMe accurately predicts DNA methylation values in whole-genome bisulfite sequencing of multiple human tissues

Cited by 6 publications

References 124 publications

Performance comparison and in-silico harmonisation of commercial platforms for DNA methylome analysis by targeted bisulfite sequencing

Performance comparison and in-silico harmonisation of commercial platforms for DNA methylome analysis by targeted bisulfite sequencing

CpG Transformer for imputation of single-cell methylomes

Accurate prediction of genome-wide RNA secondary structure profile based on extreme gradient boosting

Contact Info

Product

Resources

About