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

. 2016 Jun 9;44(13):6055–6069. doi: 10.1093/nar/gkw521

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

© The Author(s) 2016. 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 (http://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com

PMC Copyright notice

Figure 4. — Higher-order BaMMs boost accuracy of binding affinity predictions for weak sites. (A) Left: Pearson correlation between values for Klf4 binding to singly and multiply mutated consensus sites and values predicted with models trained on Klf4-bound sequences from ChIP-seq. Solid lines: BaMM models; dashed lines: PWMs from XXmotif. Middle: values measured by competitive EMSA assay versus values predicted by zeroth-order BaMM trained on ChIP-seq data. Right: same but predictions from fifth-order BaMM. Affinities to weak, multiply mutated sites (orange) are very badly predicted using a PWM (correlation 0.26) but decently using a fifth-order BaMM (correlation 0.64). (B) Left: same as (A), but showing Spearman rank correlations for FoxA2 binding affinities measured for 64 putative binding sites. DeepBind† and DeepBind differ only in model length (16 versus 24 bp). Right: Spearman correlations of our eighth-order BaMM and various other methods, adopted from Alipanahi et al. (50).

Inline graphic — Higher-order BaMMs boost accuracy of binding affinity predictions for weak sites. (A) Left: Pearson correlation between values for Klf4 binding to singly and multiply mutated consensus sites and values predicted with models trained on Klf4-bound sequences from ChIP-seq. Solid lines: BaMM models; dashed lines: PWMs from XXmotif. Middle: values measured by competitive EMSA assay versus values predicted by zeroth-order BaMM trained on ChIP-seq data. Right: same but predictions from fifth-order BaMM. Affinities to weak, multiply mutated sites (orange) are very badly predicted using a PWM (correlation 0.26) but decently using a fifth-order BaMM (correlation 0.64). (B) Left: same as (A), but showing Spearman rank correlations for FoxA2 binding affinities measured for 64 putative binding sites. DeepBind† and DeepBind differ only in model length (16 versus 24 bp). Right: Spearman correlations of our eighth-order BaMM and various other methods, adopted from Alipanahi et al. (50).

HHS Vulnerability Disclosure