Mutation Model
antigen-prime couples genetic sequence evolution with antigenic phenotype changes. Each virus carries both a nucleotide sequence and coordinates in antigenic space. Mutation events simultaneously change the sequence and may move the virus in antigenic space.
K80 Nucleotide Mutation
Mutations follow the K80 (Kimura 1980) model parameterized by a transition/transversion ratio \(\kappa\).
| Parameter | Default | Description |
|---|---|---|
transitionTransversionRatio |
5.0 | Bias toward transitions vs transversions |
Default \(\kappa = 5.0\) matches empirical observations in influenza (Rabadan et al. 2006, Bloom & Glassman 2009).
During mutation:
- A nucleotide site is randomly selected
- A new nucleotide is drawn from the K80 probability distribution
- If mutation creates a stop codon, reject and try another site
Epitope vs Non-Epitope Sites
The initial sequence encodes a protein. Users define a subset of amino-acid sites as "epitope sites" (remaining sites are "non-epitope").
| Parameter | Default | Description |
|---|---|---|
epitopeSites |
"epitopeSites.txt" | File listing epitope site indices (1-indexed) |
For influenza HA, we typically use the 49 epitope sites from Luksza & Lassig (2014).
Antigenic Space Movement
The effect of a mutation depends on whether and how it changes the protein:
| Mutation Type | Antigenic Effect |
|---|---|
| Synonymous | No movement |
| Non-synonymous at epitope site | Large movement |
| Non-synonymous at non-epitope site | Very small movement |
Step Size Distribution
Movement step sizes are drawn from gamma distributions:
| Parameter | Default | Description |
|---|---|---|
meanStepEpitope |
0.6 | Mean step size for epitope mutations (antigenic units) |
sdStepEpitope |
0.3 | Standard deviation for epitope mutations |
meanStep |
1e-5 | Mean step size for non-epitope mutations |
sdStep |
0.3 | Standard deviation for non-epitope mutations |
Default epitope step size of 0.6 antigenic units produces ~1.6 AU/year antigenic drift matching empirical HI assay data (Smith et al. 2004, Koel et al. 2013).
Step Direction
The direction \(\theta\) of movement in antigenic space is uniformly random.
| Parameter | Default | Description |
|---|---|---|
mut2D |
false | If true, allow full 360° arc; if false, 1D movement only |
Acceptance Rates
Users can apply acceptance/rejection filtering to model selection:
| Parameter | Default | Description |
|---|---|---|
epitopeAcceptance |
1.0 | Probability of accepting epitope mutations |
nonEpitopeAcceptance |
1.0 | Probability of accepting non-epitope mutations |
Setting different rates allows modeling differential selection between site types. Note: only applied to non-synonymous mutations.
High/Low Epitope Sites
For finer control, epitope sites can be subdivided into "high" and "low" categories with different step size distributions:
| Parameter | Default | Description |
|---|---|---|
proportionHighSites |
0.2 | Fraction of epitope sites designated as "high" |
meanStepEpitopeLow |
0.3 | Mean step for "low" epitope sites |
meanStepEpitopeHigh |
0.3 | Mean step for "high" epitope sites |
Predefined Vectors
By default, step sizes and directions are drawn randomly. Alternatively, users can predefine mutation vectors for each site/amino-acid pair:
| Parameter | Default | Description |
|---|---|---|
predefinedVectors |
false | Use precomputed site-specific mutation effects |
Implementation
The mutation logic is implemented in GeometricSeqPhenotype.mutate():
src/main/java/org/antigen/phenotype/GeometricSeqPhenotype.java
Key steps:
- Select random nucleotide site
- Draw mutant nucleotide from K80 distribution
- Reject if creates stop codon
- Determine if synonymous; if so, return with sequence change only
- Apply acceptance filter based on site type
- Update mutation counts (epitope/non-epitope)
- Calculate antigenic step (predefined or random)
- Return new phenotype with updated sequence and coordinates
References
- Kimura M. (1980). A simple method for estimating evolutionary rates of base substitutions. J Mol Evol.
- Smith DJ et al. (2004). Mapping the antigenic and genetic evolution of influenza virus. Science.
- Luksza M, Lassig M. (2014). A predictive fitness model for influenza. Nature.
- Koel BF et al. (2013). Substitutions near the receptor binding site determine major antigenic change during influenza virus evolution. Science.