Figure - PMC

Skip to main content

An official website of the United States government

Here's how you know

Here's how you know

Official websites use .gov
A .gov website belongs to an official government organization in the United States.

Secure .gov websites use HTTPS
A lock ( ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

View full-text article in PMC

. 2021 Sep 11;49(18):10397–10418. doi: 10.1093/nar/gkab798

Search in PMC
Search in PubMed
View in NLM Catalog
Add to search

© The Author(s) 2021. Published by Oxford University Press on behalf of Nucleic Acids Research.

This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.

PMC Copyright notice

Figure 2. — (A) The natural logarithm of the number of segments selected by StitchIt, UnifiedPeaks, and GeneHancer is shown for each dataset respectively, whereas in (B), the average length of the selected segments is depicted. (C) The number of learned models is shown, separately per consortia and method. (D) Boxplots showing Spearman correlation between predicted and measured gene-expression using linear regression with elastic net penalty considering all regions identified by StitchIt, the UnifiedPeaks approach, GeneHancer, and individual peak aggregation respectively for Blueprint and Roadmap data. Within StitchIt, UnifiedPeaks, and GeneHancer Spearman correlation was used for the initial filtering of candidate regions. Within each consortia, the same set of genes is displayed to allow comparability (Blueprint: 11140, Roadmap: 9102). As indicated by a two-sided t-test, StitchIt regions achieve the best model performance (****P ≤ 0.0001). The estimated values for the variances are: 0.018, 0.017, 0.029 (Blueprint), 0.018, 0.024, 0.032 (Roadmap), for StitchIt, UnifiedPeaks and GeneHancer, respectively. (E) The density plots delineates the number of predicted REMs per gene, shown separately for the used datasets and tested methods. Note that, due to the design of the linear model, the maximum number of predicted REMs is capped by the number of samples used for model training. (F) Considering of the entire Roadmap data set, we performed down-sampling experiments training 10 models for each gene with a different number of training samples, evaluated on the remaining of the data. According to a two-sided t-test (****P ≤ 0.0001), the performance drop is significant for each reduction of training samples.

Inline graphic — (A) The natural logarithm of the number of segments selected by StitchIt, UnifiedPeaks, and GeneHancer is shown for each dataset respectively, whereas in (B), the average length of the selected segments is depicted. (C) The number of learned models is shown, separately per consortia and method. (D) Boxplots showing Spearman correlation between predicted and measured gene-expression using linear regression with elastic net penalty considering all regions identified by StitchIt, the UnifiedPeaks approach, GeneHancer, and individual peak aggregation respectively for Blueprint and Roadmap data. Within StitchIt, UnifiedPeaks, and GeneHancer Spearman correlation was used for the initial filtering of candidate regions. Within each consortia, the same set of genes is displayed to allow comparability (Blueprint: 11140, Roadmap: 9102). As indicated by a two-sided t-test, StitchIt regions achieve the best model performance (****P ≤ 0.0001). The estimated values for the variances are: 0.018, 0.017, 0.029 (Blueprint), 0.018, 0.024, 0.032 (Roadmap), for StitchIt, UnifiedPeaks and GeneHancer, respectively. (E) The density plots delineates the number of predicted REMs per gene, shown separately for the used datasets and tested methods. Note that, due to the design of the linear model, the maximum number of predicted REMs is capped by the number of samples used for model training. (F) Considering of the entire Roadmap data set, we performed down-sampling experiments training 10 models for each gene with a different number of training samples, evaluated on the remaining of the data. According to a two-sided t-test (****P ≤ 0.0001), the performance drop is significant for each reduction of training samples.