Use DERNA Online

DERNA: Pareto Optimal mRNA Sequence Design Balancing Stability and Translation Efficiency

DERNA is an algorithm for Pareto optimal RNA design: given a target protein, it enumerates mRNA coding sequences that trade off minimum free energy (MFE) against codon adaptation index (CAI). It was developed by Xinyu Gu, Yuanyuan Qi and Mohammed El-Kebir at the University of Illinois Urbana-Champaign and published in the Journal of Computational Biology in 2024, following an earlier version at WABI 2023.

Because the genetic code is degenerate, a single protein can be encoded by an exponentially large number of mRNA sequences. These sequences differ in MFE, which relates to mRNA stability, and in CAI, which relates to translation efficiency. The two objectives usually pull against each other, so improving one tends to worsen the other.

Earlier tools commonly optimize one of these measures or a fixed combination of them. DERNA instead treats the problem as a true two-objective optimization and reports the whole set of solutions where no other sequence is better on both measures, giving designers an explicit menu of trade-offs for applications such as mRNA vaccines.

Why Pareto Optimal mRNA Design Matters for Vaccines and Therapeutics

mRNA vaccines and therapeutics depend on a sequence that is both structurally stable and efficiently translated. Since stability (MFE) and translation efficiency (CAI) conflict, a single answer hides the trade-off. A Pareto front shows every non-dominated option so that researchers can choose the balance that suits their construct.

How DERNA Traces the Pareto Front

The authors define the Pareto Optimal RNA Design problem and solve it with the weighted sum method.

1. Combine objectives: Each run optimizes a convex combination of the MFE and CAI objectives.

2. Solve by dynamic programming: Each combination is solved with dynamic programming over RNA secondary structure.

3. Sweep the weights: Varying the combination traces out the Pareto front.

DERNA Benchmark Results Against CDSfold and LinearDesign

The authors compared DERNA with CDSfold and with the heuristic version of LinearDesign (the only one publicly available) on 100 protein sequences from UniProt, plus a SARS-CoV-2 spike case study. DERNA reached the same MFE as CDSfold with higher CAI, because CDSfold ignores CAI, and its solution quality matched LinearDesign.

Spike Protein Case Study

The Pfizer-BioNTech spike mRNA has an MFE of -1217 and a CAI of 0.95 in the authors' analysis. At the same CAI, DERNA found a solution with a better MFE of -1955.2. DERNA is also fully open source, whereas key LinearDesign functionality is closed source.

What is Tamarind Bio?

Tamarind Bio is a no-code bioinformatics platform built to give life scientists and researchers access to powerful computational tools. Many cutting-edge machine learning models are hard to deploy and use. Tamarind provides an intuitive, web-based environment that removes the complexity of high-performance computing, software dependencies and command-line interfaces.

The platform is designed for biologists, chemists and other researchers who may not have a background in programming or cloud infrastructure but want to run models on their own data. Key features include:

• A user-friendly graphical interface for setting up and launching experiments

• A robust API for integration into existing research pipelines

• An automated system for managing and scaling computational resources

Tamarind treats information and data security as a top priority, as detailed in its Trust Center and Terms of Service.

Accelerating Discovery with DERNA on Tamarind Bio

• mRNA vaccine design: Generate candidate coding sequences for an antigen, as the authors demonstrate with the SARS-CoV-2 spike protein.

• Stability versus expression trade-offs: Inspect the Pareto front to choose a sequence that balances MFE and CAI.

• Method comparison: Compare DERNA designs with those from CDSfold or LinearDesign on the same target protein.

How to Use DERNA on Tamarind Bio

1. Log in and open the tool: Sign in at tamarind.bio and open DERNA.

2. Provide the target protein: Enter the amino acid sequence of the protein the mRNA should encode.

3. Set parameters: If exposed, adjust options that control the trade-off between MFE and CAI.

4. Run the job: Submit the run and wait for it to finish.

5. Download outputs: Retrieve the designed mRNA sequences from the results.

6. Interpret the front: Compare the MFE and CAI of each solution and pick the balance that suits your application.

7. Use downstream: Take chosen sequences forward for synthesis or further computational checks.

Things to Keep in Mind

• The weighted sum method has stated limitations: several k values can give the same solution, the nonconvex part of the Pareto front cannot be recovered, and sampling can be nonuniform. The authors mitigate the first by recursively examining k values.

• DERNA is slower than CDSfold and LinearDesign on the benchmark because it accounts for both MFE and CAI.

• MFE and CAI are computational proxies, not experimental measurements; the authors note other factors such as translation initiation motifs and 5-prime UTR structure also affect expression and are future work.

Source: Gu X, Qi Y, El-Kebir M. DERNA Enables Pareto Optimal RNA Design. Journal of Computational Biology 31(3), 2024. DOI: 10.1089/cmb.2023.0283. Code: https://github.com/elkebir-group/derna.