(423e) Exploiting Thermodynamics of mRNA Secondary Structure for Antisense Design

The ability to manipulate gene expression is one of the most important techniques in biomolecular engineering, with potential benefits spanning a wide spectrum of fields from gene therapy to metabolic engineering. One specific method of gene regulation, which has proven to be effective in natural and artificial systems alike, is the use of antisense RNA to inhibit protein synthesis. Despite biotechnological advances, the design of effective antisense sequences remains a formidable task. In an effort to remedy this issue, a novel method of predicting effective antisense is proposed. In recent years, the idea that an mRNA strand may not always take the form of a distinct fixed molecular structure has become much more prominent. It is believed that an mRNA molecule may actually be in a state of constant structural fluctuation, transitioning between different conformations near the minimum free energy (MFE) structure, particularly in an ever-changing cellular environment. Analyzing suboptimal mRNA structures with a thermodynamic stability comparable to that of the MFE structure may reveal that certain regions are more “volatile” than others. Since these regions have the ability to change conformation without significantly altering the Gibb’s free energy of the entire molecule, they may have more freedom to alter their hydrogen bonding. Therefore, these regions would likely be the most accessible targets for antisense binding because of their incessant making and breaking of intramolecular hydrogen bonds. A computational framework, GenAVERT, was developed using the existing programs, UNAFold and RNAforester. UNAFold is used to predict the minimum free energy structure and a set of suboptimal structures, which are then structurally compared to each other using RNAforester, revealing regions of similarity and dissimilarity. Regions of dissimilarity are proposed to indicate sites in an mRNA structure that would be particularly accessible for antisense binding since they display a theoretically lower level of thermodynamic stability over the suboptimal range of free energy. Validation was carried out by predicting antisense sequences and comparing results to actual sequences found in natural toxin-antitoxin antisense systems of bacteria. GenAVERT predicted that the most volatile regions of those toxin-encoding mRNAs were essentially the same as those of the natural antisense target sites. The significance of these toxin-antitoxin systems is in the necessity for efficient inhibition of protein expression to avoid cell death. Translation levels must be brought to an extremely low level or blocked completely for a cell to continue to carry these suicide genes, therefore their corresponding antisense inhibitors must be exceedingly effective. A successful antisense prediction system should be able to predict sequences similar to these RNA antitoxins after analyzing their correlating toxin-encoding mRNAs. As a benchmark, other currently available programs designed for the development of inhibitory RNA were also used to predict potential antisense for the same mRNA sequences and compared. The results revealed that predicting antisense sequences using the concept of volatility in mRNA secondary structure provided a much more accurate prediction method without requiring that the user have detailed knowledge of rational antisense design.