GenAIRR 0.4.1

An advanced immunoglobulin sequence simulation suite for benchmarking alignment models and sequence analysis.

GenAIRR

Adaptive Immune Receptor Repertoire sequence simulator
Generate realistic BCR & TCR repertoires in a single line of Python.

---

📑 Table of Contents

1. Why GenAIRR?
2. Key Features
3. Installation
4. Quick Start
5. Examples
6. Mutation Models
7. Roadmap
8. Contributing
9. Citing GenAIRR
10. License
11. Acknowledgements

---

🧐 Why GenAIRR?

Click to expand

Benchmarking modern aligners, exploring somatic-hypermutation, or stress-testing novel ML pipelines requires large, perfectly-annotated repertoires—not snippets of real data peppered with sequencing error.
GenAIRR fills that gap with a plug-and-play, fully-extensible simulation engine that produces sequences while giving you full ground-truth labels.

---

✨ Key Features

Category	Highlights
Realistic Simulation	Context-aware S5F mutations, indels, allele-specific trimming, NP-region modelling
Composable Pipelines	Chain together built-in & custom AugmentationSteps into simulation pipelines
Multi-Chain Support	Heavy & light BCRs plus TCR-β out of the box
Research-ready Output	JSON / pandas export, built-in plotting stubs, deterministic seeds
Docs & Tutorials	Rich API docs, Jupyter notebooks, step-by-step guides

---

⚡ Installation

# Python ≥ 3.9
pip install GenAIRR
# or the bleeding edge
pip install git+https://github.com/your-org/GenAIRR.git

---

🚀 Quick Start

Below is a 60-second tour. See /examples for notebooks and CLI usages.

from GenAIRR.pipeline import AugmentationPipeline
from GenAIRR.parameters import ChainType,CHAIN_TYPE_INFO
from GenAIRR.steps import SimulateSequence, FixVPositionAfterTrimmingIndexAmbiguity
from GenAIRR.mutation import S5F
from GenAIRR.data import builtin_heavy_chain_data_config
from GenAIRR.steps.StepBase import AugmentationStep

# 1️⃣  Configure built-in germline & chain type
data_cfg = builtin_heavy_chain_data_config()
AugmentationStep.set_dataconfig(config = data_cfg,
                                chain_type=ChainType.BCR_HEAVY)

# 2️⃣  Build a minimal pipeline
pipeline = AugmentationPipeline([
    SimulateSequence(mutation_model=S5F(min_mutation_rate=0.003, max_mutation_rate=0.25), productive=True),
    FixVPositionAfterTrimmingIndexAmbiguity()
])

# 3️⃣  Simulate!
sim = pipeline.execute()
print(sim.get_dict())

---

🧑‍💻 Examples

1. Full Heavy-Chain Pipeline

from GenAIRR.steps import (
    FixDPositionAfterTrimmingIndexAmbiguity, FixJPositionAfterTrimmingIndexAmbiguity,
    CorrectForVEndCut, CorrectForDTrims, CorruptSequenceBeginning,
    InsertNs, InsertIndels, ShortDValidation, DistillMutationRate
)

pipeline = AugmentationPipeline([
    SimulateSequence(mutation_model=S5F(min_mutation_rate=0.003, max_mutation_rate=0.25), productive=True),
    FixVPositionAfterTrimmingIndexAmbiguity(),
    FixDPositionAfterTrimmingIndexAmbiguity(),
    FixJPositionAfterTrimmingIndexAmbiguity(),
    CorrectForVEndCut(),
    CorrectForDTrims(),
    CorruptSequenceBeginning(
        corruption_probability=0.7,
        corrupt_events_proba=[0.4, 0.4, 0.2],
        max_sequence_length=576,
        nucleotide_add_coefficient=210,
        nucleotide_remove_coefficient=310,
        nucleotide_add_after_remove_coefficient=50,
        random_sequence_add_proba=1
    ),
    InsertNs(n_ratio=0.02, proba=0.5),
    ShortDValidation(short_d_length=5),
    InsertIndels(indel_probability=0.5, max_indels=5, insertion_proba=0.5, deletion_proba=0.5),
    DistillMutationRate()
])
result = pipeline.execute()

2. Naïve Sequence (no SHM)

from GenAIRR.mutation import Uniform
naive_step = SimulateSequence(mutation_model=Uniform(0, 0), productive=True)
pipeline = AugmentationPipeline([naive_step])
naive_seq = pipeline.execute()
print(naive_seq.sequence)

3. Custom Allele Combination

custom_step = SimulateSequence(
    mutation_model=S5F(0.003, 0.25),
    productive=True,
    specific_v=data_cfg.allele_list('v')[0],# specific V allele (as Allele object)
    specific_d=data_cfg.allele_list('d')[0],# specific D allele (as Allele object)
    specific_j=data_cfg.allele_list('j')[0] # specific J allele (as Allele object)
)
pipeline = AugmentationPipeline([custom_step])
print(pipeline.execute().get_dict())

---

🔬 Mutation Models

Model	Description	When to use
S5F	Context-specific somatic hyper-mutation	Antibody maturation studies
Uniform	Evenly random mutations	Baselines / ablation
Your Model ➕	Implement BaseMutationModel	Custom evolutionary scenarios

from GenAIRR.mutation import S5F
s5f = S5F(min_mutation_rate=0.01, max_mutation_rate=0.05)
mut_seq, muts, rate = s5f.apply_mutation(naive_seq)

---

🗺️ Roadmap

• 🚧 More Complex Mutation Model (With Selection)
• 🚧 More Built-in Data Configs (e.g., TCR, custom germlines)
• 🚧 More Built-in Steps (e.g., more mutation models, more data augmentation)
• 🚧 Deeper Docs (e.g., more examples, more tutorials)

See the open issues. Feel something’s missing? Open a feature request.

---

🤝 Contributing

Contributions are welcome! 💙 Please read our contributing guide and check the good first issue label.

---

✏️ Citing GenAIRR

If GenAIRR helps your research, please cite:

Konstantinovsky T, Peres A, Polak P, Yaari G.  
An unbiased comparison of immunoglobulin sequence aligners.
Briefings in Bioinformatics. 2024 Sep 23; 25(6): bbae556.  
https://doi.org/10.1093/bib/bbae556  
PMID: 39489605 | PMCID: PMC11531861

---

📜 License

Distributed under the MIT License. See LICENSE for details.

---

🙏 Acknowledgements

GenAIRR is inspired by and builds upon amazing work from the immunoinformatics community—especially AIRRship.