To get mRNA therapeutics to the clinic quickly, drug developers need a successful RNA design. Starting with a well-designed RNA candidate that considers the entire RNA lifecycle helps prevent manufacturing and quality control issues later in the development pipeline, saving drug developers both time and money. Strong designs also improve the efficacy of the therapeutic, providing patients with effective treatments. 

Often drug developers rely on codon optimization to design functional, successful RNA. However, codon optimization is only one aspect of a thoughtful design process and focuses on a single factor of the RNA, the coding sequence itself, and ignores the other elements of the therapeutic. 

For a truly comprehensive RNA candidate that is both manufacturable and therapeutically effective, drug developers need to design with a multi-objective approach. 

Codon optimization: Addressing the common design method 

Codons are three-nucleotide-long segments of mRNA that code for a specific amino acid. During translation, a ribosome reads the codons on the mRNA and calls upon a tRNA to gather the amino acids that correspond with the codons. Amino acids are the building blocks of proteins, impacting every cellular function.  

There are 64 codons: the 4 nucleotides arranged in every possible 3-nucleotide segment (4³ combinations). However, there are only 20 amino acids, leading to multiple codons coding for the same amino acid. This concept is called codon degeneracy, and it adds protection against point mutations when a single nucleotide is swapped. For example, the amino acid arginine can be coded by AGA, AGG, and CGN, where N is any of the four nucleotides. If there is a point mutation in the third position of a CGA codon, then the mutated codon will still code for arginine since it will still be a degenerate CGN codon. 

Drug developers can use codon degeneracy when designing RNA through codon optimization. While multiple codons may code for the same amino acid, certain degenerate codons may be better or worse for translation in a given cell type. Drug developers can consider this and design RNA sequences that optimize certain degenerate codons, typically the most abundant codon in the cell. However, many drug developers rely too heavily on codon optimization for RNA design at the expense of other features. Codon optimization only considers the coding sequence of the RNA, forgoing other variables like 5′ and 3′ UTRs and poly(A) tails. Performing solely codon optimization for RNA design leaves additional design improvements on the table, resulting in an RNA candidate that does not fully consider stability, quality control, or even optimized therapeutic function. 

Improved RNA design: A multi-objective approach 

For truly optimized RNA candidates, drug developers must consider all variables and performance dimensions of the RNA. Focusing entirely on a single performance dimension will worsen others, leading to a skewed RNA candidate that may not meet regulatory standards. An RNA that has only undergone codon optimization to speed up translation may not be stable enough to be delivered in a therapeutic, for example. 

In multi-objective RNA design, drug developers consider tradeoffs in performance dimensions, deciding what level of each dimension is most valuable for their RNA candidate. It’s impossible to have an RNA that simultaneously has the fastest expression and highest stability. Identifying the right ratio of these and other performance dimensions, including IVT yield, sustained expression, and integrity, is key to a design optimized to a therapeutic's Target Product Profile. 

This process is not linear. Rather, it involves drug developers understanding what performance dimensions they want to target and evaluating the design solutions and tradeoffs that get the RNA to that target. 

AI and machine learning models can help evaluate these design choices. Models trained on modified therapeutic mRNA have learned how different design elements impact RNA candidates. These models can generate effective sequences. Yet, many models do not have the proper training data to do this. These models are trained using unmodified mRNA, which does not provide the data needed to generate sequences with a multi-objective therapeutic goal in mind. To design clinic-ready RNA, models must be trained on therapeutic RNA. 

Sequence design at Eclipsebio 

At Eclipsebio, we perform multi-objective sequence design in our eNAVIGATE platform. With eNAVIGATE, we consider how design changes impact an RNA therapeutic throughout its lifecycle, not just during manufacturing. We use AI-powered design models that are continuously trained on modified therapeutic mRNA to evaluate tradeoffs on performance dimensions based on multiple design variables. 

Beyond models, each partner's mRNA is designed against their Target Product Profile, not a generic benchmark. Whether a partner needs a certain on-target editing efficiency, durable expression over a defined window, a tight immunogenicity ceiling, or a combination of factors, we can help reach that goal. We provide generated designs as a curated set spanning the relevant tradeoffs, so partners can choose against their goals rather than against a generic default. 

eNAVIGATE is a part of our lab-in-the-loop eCOMPASS platform that combines AI-powered design with our eMERGE platform’s sequencing-based characterization. eMERGE provides quality control insights into an RNA candidate, revealing where ribosomes are stalling, how the RNA is folding, and where degradation is occurring. These insights not only inform how to proceed with the current design, but they also are integrated into our eNAVIGATE models to guide future RNA designs. 

Interested in how eNAVIGATE can optimize your sequence? Contact Eclipsebio to learn more.