Fig. 3: Experimental validation of iM6A modeling.
From: Deep learning modeling m6A deposition reveals the importance of downstream cis-element sequences
a The workflow of the validation of iM6A by m6A-QTLs dataset (See details in Methods). b The change direction of experimentally measured m6A peak changes agreed with that of the iM6A modeling on the related SNVs. The expected value was 23.5 in random situation, and the observed value was 33. The p-value (<0.004) was calculated based on Binomial distribution. c The change quantity of experimentally measured m6A peak changes (ΔPeakEnrichment) agreed with that of the iM6A modeling on the related SNVs. The R-value was calculated by Pearson Correlation Coefficient, and p-value (<0.0003) was determined by two-sided Student’s t-test. d Bar plot of SNVs distribution over the corresponding m6A site. 10, 4, and 33 SNVs were located upstream, at, or downstream of the corresponding m6A site respectively. The p-value (<0.0002) was calculated based on Binomial distribution. e The change quantity of pentamer effective value changes (ΔEffectiveValue) agreed with experimentally measured m6A peak changes (ΔPeakEnrichment) on the related SNVs. The R-value was calculated by Pearson Correlation Coefficient, and p-value (<0.001) was determined by two-sided Student’s t-test. f–i Examples of SNVs affecting m6A deposition. rs7831 (Fig. 2f), rs75907001 (Fig. 2g) abolished or dampened m6A deposition respectively. rs9090 (Fig. 2h) and rs1057278 (Fig. 2i) enhanced m6A deposition. The m6A site was marked by red color, the mutation site was marked by blue color. The enhancers and silencers were marked by red and green lines, respectively. The m6A RIP-seq data for homozygote of the major allele, heterozygote, homozygote of the minor allele, and Input were marked by red, green, blue, and black color, respectively. The dot plot showed the probability value of the m6A with major allele or minor allele.
