Limits...
Role of 3'UTRs in the translation of mRNAs regulated by oncogenic eIF4E--a computational inference.

Santhanam AN, Bindewald E, Rajasekhar VK, Larsson O, Sonenberg N, Colburn NH, Shapiro BA - PLoS ONE (2009)

Bottom Line: Interesting tendencies of secondary structure stability are found near the start codon and at the beginning of the 3'UTR region.Highly upregulated mRNAs show negative selection (site avoidance) for binding sites of several microRNAs.These results are consistent with the emerging model of regulation of mRNA translation through a dynamic balance between translation initiation at the 5'UTR and microRNA binding at the 3'UTR.

View Article: PubMed Central - PubMed

Affiliation: Gene Regulation Section, Laboratory of Cancer Prevention, National Cancer Institute, Frederick, Maryland, United States of America.

ABSTRACT
Eukaryotic cap-dependent mRNA translation is mediated by the initiation factor eIF4E, which binds mRNAs and stimulates efficient translation initiation. eIF4E is often overexpressed in human cancers. To elucidate the molecular signature of eIF4E target mRNAs, we analyzed sequence and structural properties of two independently derived polyribosome recruited mRNA datasets. These datasets originate from studies of mRNAs that are actively being translated in response to cells over-expressing eIF4E or cells with an activated oncogenic AKT: eIF4E signaling pathway, respectively. Comparison of eIF4E target mRNAs to mRNAs insensitive to eIF4E-regulation has revealed surprising features in mRNA secondary structure, length and microRNA-binding properties. Fold-changes (the relative change in recruitment of an mRNA to actively translating polyribosomal complexes in response to eIF4E overexpression or AKT upregulation) are positively correlated with mRNA G+C content and negatively correlated with total and 3'UTR length of the mRNAs. A machine learning approach for predicting the fold change was created. Interesting tendencies of secondary structure stability are found near the start codon and at the beginning of the 3'UTR region. Highly upregulated mRNAs show negative selection (site avoidance) for binding sites of several microRNAs. These results are consistent with the emerging model of regulation of mRNA translation through a dynamic balance between translation initiation at the 5'UTR and microRNA binding at the 3'UTR.

Show MeSH

Related in: MedlinePlus

Support vector machine classifier effectively predicts fold change.Log-log plot of the eIF4E dataset fold change plotted with the corresponding support vector machine classifier results. The used eIF4E overexpression dataset consists of 4000 mRNAs for training and 5629 mRNAs for testing the classifier (see Materials and Methods).
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC2654073&req=5

pone-0004868-g003: Support vector machine classifier effectively predicts fold change.Log-log plot of the eIF4E dataset fold change plotted with the corresponding support vector machine classifier results. The used eIF4E overexpression dataset consists of 4000 mRNAs for training and 5629 mRNAs for testing the classifier (see Materials and Methods).

Mentions: As described in the Methods section, we used the libsvm package to train a support vector machine with features of the mRNA: total length, 3′UTR length and G+C content. The support vector machine was used to predict a real-valued number instead of a Boolean two-class classification result. Using these features, the support vector machine prediction compared with the logarithm of the fold changes yielded a Spearman correlation coefficient of 0.55 (Table 3 and Figure 3). Note that the use of the support vector machine (Table 3) leads to higher correlation coefficients compared to the results of the individual features (Tables 1 and 2). The prediction accuracy without the G+C content feature has a correlation coefficient of 0.43. Applying the support vector machine trained on the eIF4E overexpression dataset to the AKT activation dataset, we found that the support vector machine can predict the AKT fold change with a Spearman correlation coefficient of 0.15. This suggests commonalities (as well as differences) between genes differentially expressed due to eIF4E overexpression or AKT activation. 79% percent of the mRNAs that shifted in response to eIF4E were predicted, while 62% percent of those that did not shift were predicted. This result indicates a high degree of accuracy for the support vector machine (Table 4).


Role of 3'UTRs in the translation of mRNAs regulated by oncogenic eIF4E--a computational inference.

Santhanam AN, Bindewald E, Rajasekhar VK, Larsson O, Sonenberg N, Colburn NH, Shapiro BA - PLoS ONE (2009)

Support vector machine classifier effectively predicts fold change.Log-log plot of the eIF4E dataset fold change plotted with the corresponding support vector machine classifier results. The used eIF4E overexpression dataset consists of 4000 mRNAs for training and 5629 mRNAs for testing the classifier (see Materials and Methods).
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC2654073&req=5

pone-0004868-g003: Support vector machine classifier effectively predicts fold change.Log-log plot of the eIF4E dataset fold change plotted with the corresponding support vector machine classifier results. The used eIF4E overexpression dataset consists of 4000 mRNAs for training and 5629 mRNAs for testing the classifier (see Materials and Methods).
Mentions: As described in the Methods section, we used the libsvm package to train a support vector machine with features of the mRNA: total length, 3′UTR length and G+C content. The support vector machine was used to predict a real-valued number instead of a Boolean two-class classification result. Using these features, the support vector machine prediction compared with the logarithm of the fold changes yielded a Spearman correlation coefficient of 0.55 (Table 3 and Figure 3). Note that the use of the support vector machine (Table 3) leads to higher correlation coefficients compared to the results of the individual features (Tables 1 and 2). The prediction accuracy without the G+C content feature has a correlation coefficient of 0.43. Applying the support vector machine trained on the eIF4E overexpression dataset to the AKT activation dataset, we found that the support vector machine can predict the AKT fold change with a Spearman correlation coefficient of 0.15. This suggests commonalities (as well as differences) between genes differentially expressed due to eIF4E overexpression or AKT activation. 79% percent of the mRNAs that shifted in response to eIF4E were predicted, while 62% percent of those that did not shift were predicted. This result indicates a high degree of accuracy for the support vector machine (Table 4).

Bottom Line: Interesting tendencies of secondary structure stability are found near the start codon and at the beginning of the 3'UTR region.Highly upregulated mRNAs show negative selection (site avoidance) for binding sites of several microRNAs.These results are consistent with the emerging model of regulation of mRNA translation through a dynamic balance between translation initiation at the 5'UTR and microRNA binding at the 3'UTR.

View Article: PubMed Central - PubMed

Affiliation: Gene Regulation Section, Laboratory of Cancer Prevention, National Cancer Institute, Frederick, Maryland, United States of America.

ABSTRACT
Eukaryotic cap-dependent mRNA translation is mediated by the initiation factor eIF4E, which binds mRNAs and stimulates efficient translation initiation. eIF4E is often overexpressed in human cancers. To elucidate the molecular signature of eIF4E target mRNAs, we analyzed sequence and structural properties of two independently derived polyribosome recruited mRNA datasets. These datasets originate from studies of mRNAs that are actively being translated in response to cells over-expressing eIF4E or cells with an activated oncogenic AKT: eIF4E signaling pathway, respectively. Comparison of eIF4E target mRNAs to mRNAs insensitive to eIF4E-regulation has revealed surprising features in mRNA secondary structure, length and microRNA-binding properties. Fold-changes (the relative change in recruitment of an mRNA to actively translating polyribosomal complexes in response to eIF4E overexpression or AKT upregulation) are positively correlated with mRNA G+C content and negatively correlated with total and 3'UTR length of the mRNAs. A machine learning approach for predicting the fold change was created. Interesting tendencies of secondary structure stability are found near the start codon and at the beginning of the 3'UTR region. Highly upregulated mRNAs show negative selection (site avoidance) for binding sites of several microRNAs. These results are consistent with the emerging model of regulation of mRNA translation through a dynamic balance between translation initiation at the 5'UTR and microRNA binding at the 3'UTR.

Show MeSH
Related in: MedlinePlus