Design tools for Synthetic Virology

1. Design tools for Synthetic Virology Dimitris Papamichail Assistant Professor Computer Science Department University of Miami

2. Our group I have been working on design and synthesis of genomic sequences in collaboration with: Professor Steven Skiena Computer Science Department Stony Brook University Professor Eckard Wimmer Department of Molecular Genetics and Microbiology Stony Brook University Other members: Rob Coleman, Steffen Mueller, Bruce Futcher

3. Initial motive: Vaccine Design Question: How can we rapidly create a vaccine for a new viral disease? Motivation: Mutated lethal viruses, bio-terrorism, cheap synthesis technologies Input: The genome of a virus Output: Our design of a "better" virus to serve as a vaccine candidate Aim: Redesign life  (if you assume viruses are alive)

4. Our designs So far, our group has designed, synthesized, and evaluated (or still evaluates): Four new variants of poliovirus - Two codon optimized designs - Two codon pair optimized designs Several optimized flu virus segments Recently a couple of bacterial genes A couple of artificial constructs for hypothesis testing.

5. Novel sequence design Our goal • Debilitate virus genome translation / replication. • Make a difficult to revert and genetically stable virus as the cumulative phenotype of many mutations each with a small effect. • Optimization problem at hand: • Select one (or few) of the 7.9 x 10442 possible synonymous encodings for a poliovirus capsid gene of 881 amino acids (compared to an estimated 1.3 x 1079 atoms in the known universe), that serves our design criteria.

6. Novel sequence design First two designs: Codon (de-)optimized • We tested the hypothesis that underrepresented codons reduce translation efficiency by creating a novel polio capsid design (PV-AB) which: • Encoded the same amino-acid sequence • Used only the least frequent codon for each amino-acid in human brain specific genes (and in human tissues in general). Total number of silent mutations: 680 • We also created another design (PV-SD) which maximized the hamming distance of the capsid encoding, while keeping the same codon frequency distribution. • Total number of silent mutations: 934 • Why alter only the capsid coding region? No cis-acting structural RNA elements

7. Novel sequence design Experiments • The “shuffled” polio design translates relatively well and is as potent in killing mice as the wildtype. • The brain-hostile design translates minimally, but use of smaller segments leads to attenuated strains.

8. Novel sequence design Methods • To achieve maximum hamming distance without altering the codon distribution, we used maximum weight bipartite matching between codon positions and codons, using as weight the number of bases changed. • Restriction sites were inserted uniquely (inserted in specific areas and then eliminated everywhere else). • Certain regions were locked to preserve secondary structure.

9. Novel sequence design Codon pair bias • According to Hatfield et al., another source of translation (in)efficiency is the codon pair bias. • We quantified the bias with the following score: • Many viruses actually are using overrepresented codon pairs (in human) to encode their genes.

10. Novel sequence design Codon pair bias • There also seems to be significant correlation between related eucaryotes and codon pair bias.

11. Novel sequence design Another two designs: Codon pair (de-)optimized • We designed two polio capsid sequences that optimize the usage of over-represented and under-represented codon pairs in human.

12. Novel sequence design Our designs • The design consisted of the following steps: • Same codon frequency distribution • Optimized codon pair score • Restriction site uniqueness and elimination • Local Secondary structure folding energy restriction • Splice site elimination • Goals achieved with simulated annealing, optimization passes and manual intervention.

13. Novel sequence design Codon pair optimization problem • Our problem: variant of TSP (Travelling Salesperson Problem) • This variant is polynomially time solvable, but with O(n65) complexity, using dynamic programming. • We have 20 countries (amino-acids), 64 cities (codons), each country has from 1 to 6 cities. • We will make n visits to the countries (amino-acid chain). • Each time we visit a country we can visit only one city (select the codon to code for an amino acid). • The total number of times we visit each city is fixed (codon distribution).

14. Novel sequence design Results How do the new codon pair biased designs behave? • maxP1 (using overrepresented codon pairs, 566 mutations) translates as well as the wildtype. • minP1 (using underrepresented codon pairs, 631 mutations) translates poorly. • Details on the results can be found in our June 27, 2008 Science publication Currently we are also investigating other signals, such as CpG dinucleotide content, which are inherent in such biased constructs.

15. Sequence design tools • Tools already exist in sequence design: • GeMS • CAD-PAM • Gene2oligo • DNAWorks • GeneDesign • GenoCAD • Most of these tools offer: • Oligo-design • Restriction site creation/elimination • Codon usage alteration • Oligo-design is the process of breaking the sequence into oligos (short sequence fragments), which can be used to self-assemble through PCR cycles into synthons (or the whole sequence, if small enough). • We consider this process as a black box provided by synthesis companies.

16. Our goal • Efficient Algorithm Design for coding sequence alterations. • Techniques for multi-objective optimization in genomic sequence design. • Expand the knowledge base of systematic sequence bias in genomic sequences • Incorporation of pathways, transcriptional control, boolean logic and other complex criteria in novel genomic sequence design. We aim to create a system conceptually built around constraints instead of sequences. The gene/genome designer will work on the level of specifying characteristics of the desired gene/genome (amino acid sequences, codon/codon-pair distribution, distribution of restriction sites, RNA secondary structure constraints, incorporation/elimination of patterns, etc.) and the gene editor will algorithmically design a DNA sequence realizing these constraints.

17. Sequence design tools SeEd (Sequence Editor)

18. Thank you!

19. Questions