wavess 1.2: Presenting an HLA-aware within-host virus sequence simulation framework

The paper introduces wavess 1.2, an enhanced within-host virus sequence simulation framework that incorporates HLA-aware cytotoxic T-lymphocyte responses and variable recombination rates to more accurately model immune-driven viral evolution and diversification.

The Gist

Imagine your body is a bustling city, and a virus is an intruder trying to break in and take over. For a long time, scientists have tried to build computer simulations to predict how this intruder evolves and changes its "costume" to hide from the city's security forces.

This paper introduces an upgrade to a specific computer program called wavess (version 1.2). Think of wavess as a high-tech video game engine that simulates a virus infection inside a single person. The authors, Zena Lapp and Thomas Leitner, have added two major new features to make the game much more realistic.

Here is a breakdown of what they did, using simple analogies:

1. The "Smart" Security Guard (The CTL Response)

The Old Way:
In the previous version of the game, the virus had to dodge a generic "immune system." It was like the virus was running from a foggy, blurry security guard who didn't know exactly what the virus looked like. The virus could hide, but the simulation didn't capture the specific, sharp attacks that happen in real life.

The New Upgrade:
The new version adds HLA-aware CTLs (Cytotoxic T-Lymphocytes).

• The Analogy: Imagine the security guard now has a specific "Wanted Poster" for every single criminal in the city. These posters are unique to every person (based on their genetics, called HLA).
• How it works: The virus has a specific "uniform" (a peptide). If the virus's uniform matches the "Wanted Poster" on the guard's clipboard, the guard attacks immediately.
• The Escape: To survive, the virus must change its uniform just enough so it no longer matches the poster. This is called an "escape mutation."
• The Catch: Changing the uniform is risky. If the virus changes too much, it might break its own engine (fitness cost). The simulation now calculates this delicate balance: How much can the virus change to hide, without breaking itself?

2. The "Shuffling Deck" (Variable Recombination)

The Old Way:
Viruses sometimes swap pieces of their genetic code, like shuffling two decks of cards together to create a new hand. In the old simulation, this shuffling happened at a steady, boring rate everywhere in the genome, like shuffling a deck of cards where every card has an equal chance of being swapped.

The New Upgrade:
The new version allows for variable recombination rates.

• The Analogy: Imagine the virus genome is a long train with different carriages. Some carriages are glued together tightly (hard to swap), while others are on a slippery track (easy to swap).
• Hotspots: The simulation can now create "hotspots" where shuffling happens constantly, and "cold spots" where it rarely happens.
• Why it matters: This is crucial for viruses that have separate sections (like a segmented genome) or for viruses where different genes are far apart. It allows the simulation to mimic real-world scenarios where the virus swaps entire "carriages" rather than just single cards.

The Real-World Test: The HIV Example

To prove their new game engine works, the authors ran a simulation using HIV-1.

• They set up a "battlefield" with two specific genes: pol (the engine) and gp120 (the disguise).
• They gave the virus a specific set of "Wanted Posters" (HLA types) that a real human might have.
• The Result: The simulation showed that the virus eventually changed its disguise to escape the guards.
  • If the human had many "Wanted Posters" (many epitopes), the virus took longer to escape, and the outcome was more unpredictable.
  • If the human had fewer posters, the virus escaped quickly and consistently.
  • The simulation also correctly identified where the virus swapped its genetic "train cars" (recombination), matching what we see in real data.

Why Should You Care?

This isn't just a video game update; it's a tool for saving lives.

1. Vaccine Design: By understanding exactly how a virus changes its "uniform" to escape specific security guards (HLA types), scientists can design vaccines that cover more of these variations.
2. Tracking Outbreaks: It helps researchers understand how viruses spread and evolve in different populations, which is vital for stopping pandemics.
3. Personalized Medicine: Since everyone has different "Wanted Posters" (HLA types), this tool helps predict how a specific person's body might fight a specific virus.

In short: The authors upgraded their virus simulation from a generic "hide and seek" game to a high-stakes, strategic battle where the virus must constantly change its disguise to survive a smart, personalized security force, all while shuffling its genetic deck in realistic ways. This helps us build better defenses against future viral threats.

Technical Summary

1. Problem Statement

Understanding how viral sequences evolve under selective pressure is critical for vaccine design and inferring transmission networks. A major selective pressure is the CD8+ cytotoxic T-lymphocyte (CTL) response, which targets viral peptides presented by host Human Leukocyte Antigen (HLA) class I molecules. Viruses often escape this pressure by mutating specific amino acids, leaving distinct "escape signatures" in their genomes.

However, existing within-host virus simulation frameworks (e.g., SLiM, fwdpy11) generally lack the ability to explicitly model HLA-specific CTL responses. They often treat immune pressure as a generic adaptive response or fail to account for the variability introduced by different host HLA genotypes. Additionally, previous versions of the authors' own simulator, wavess, did not support variable recombination rates, limiting the modeling of recombination hotspots, non-adjacent genes, or segmented genomes.

2. Methodology

The authors updated their existing individual-based, forward-time simulation framework, wavess, to version 1.2. The core of the tool is written in Python 3 with R helper functions. The key methodological updates include:

A. HLA-Aware CTL Immune Response

• Mechanism: The model now separates the B-cell (antibody) response from the T-cell (CTL) response.
• Epitope Recognition: The system identifies T-cell epitopes based on a founder virus sequence and a specific set of host HLA alleles (HLA-A and HLA-B).
• Escape Dynamics:
  • A recognized epitope imposes a fitness cost that increases linearly over time until a maximum cost ($c_{max}$) is reached.
  • Complete Escape: A single mutation at a specific "anchor" position (or any position within the epitope, depending on configuration) confers immediate and complete escape (fitness = 1).
  • Fitness Calculation: The total T-cell fitness ($F_{IT}$) for a virus is the product of the fitness of all $m$ recognized epitopes. The overall viral fitness is the product of conserved fitness, replicative fitness, B-cell fitness, and T-cell fitness.
• Parameters: Users can define the time to reach maximum fitness cost ($t_{max}$) for each epitope, allowing for the simulation of varying maturation rates of the immune response.

B. Variable Recombination Rates

• Mechanism: Unlike version 1.0, which used a uniform recombination rate, version 1.2 allows for position-specific recombination rates.
• Implementation:
  • The recombination rate is converted into a probability of an odd number of recombination events occurring at a breakpoint, calculated as $(1 - e^{-2n\lambda})/2$ to account for the Poisson process of recombination events.
  • Optimization: To handle computational efficiency, the algorithm uses binomial sampling for positions with a baseline rate (if >95% of positions share the same rate) and individual sampling for positions with distinct rates (e.g., recombination hotspots or gene boundaries).
• Applications: This enables the modeling of:
  1. Recombination hotspots.
  2. Non-adjacent genes (high recombination rates between distant loci).
  3. Segmented genomes (reassortment rates).

C. Validation Case Study (HIV-1)

• Setup: The authors simulated the evolution of HIV-1 genes pol (polymerase) and gp120 (envelope) over one year.
• Inputs:
  • Founder Sequence: Subtype B (DEMB11US006).
  • HLA Panel: 27 alleles from the IEDB panel (combinations of 2 HLA-A and 2 HLA-B).
  • Epitope Prediction: Used NetMHCpan 4.1 to predict 9-mer epitopes with specific immunogenicity thresholds.
  • Recombination: Set a baseline rate of $1.5 \times 10^{-5}$ within genes and a higher rate ($0.17$) between pol and gp120 to simulate splicing boundaries.
• Analysis: Used 3seq to detect recombination breakpoints in simulated sequences and tracked mean CTL fitness over time.

3. Key Contributions

1. First HLA-Aware Within-Host Simulator: wavess 1.2 is the first framework to explicitly model the interaction between specific host HLA alleles and viral epitopes, allowing for the simulation of escape mutations driven by CTL pressure.
2. Flexible Recombination Modeling: The introduction of variable recombination rates allows for biologically realistic modeling of complex genomic architectures, including hotspots and segmented viruses.
3. Open-Source Implementation: The updated code and vignettes are publicly available on GitHub, facilitating reproducibility and extension by the research community.

4. Results

• CTL Escape Dynamics:
  • Simulations showed that the time required for a virus to achieve high mean CTL fitness (>0.9, indicating escape) varied significantly based on the number of recognized epitopes.
  • Correlation: There was a moderate positive correlation (Spearman $\rho = 0.49$) between the number of epitopes and the time to escape.
  • Variability: Viruses with fewer epitopes escaped more quickly and with less stochastic variability (SD = 10 days for 2 epitopes) compared to those with more epitopes (SD = 37 days for 12 epitopes).
  • Timeframe: Escape times ranged from 35 days to over 365 days, aligning with empirical literature on HIV escape (weeks to years).
• Recombination Detection:
  • The 3seq tool successfully identified recombination breakpoints in the simulated data.
  • Breakpoints were frequently detected at the junction where pol and gp120 were spliced, validating the model's ability to simulate high recombination rates between non-adjacent genes.
  • Recombination events were also detected throughout the genes, consistent with the baseline rate.

5. Significance

• Vaccine Design: By accurately simulating how CTLs drive viral evolution, researchers can better predict which viral regions are likely to mutate under immune pressure, aiding in the design of vaccines that target conserved, essential epitopes.
• Transmission Inference: The framework improves the accuracy of inferring transmission networks by accounting for the specific selective pressures (HLA types) that shape viral sequences in the donor and recipient.
• Beyond Virology: While demonstrated on HIV, the HLA-aware framework and variable recombination capabilities are applicable to other pathogens and even non-viral systems where host-specific immune selection and complex recombination occur.
• Mechanistic Insight: The tool provides a platform to investigate how specific biological mechanisms (e.g., epitope immunogenicity, recombination hotspots) influence the trajectory of within-host evolution.