pepflex

PepFlex

Python framework for in silico peptide evolution: random generation, mutation/crossover, custom fitness evaluation, and multi-round population optimization.

Repo: github.com/Kdevos12/PepFlex | PyPI: pepflex==0.0.4

When to Use This Skill

• Running evolutionary / genetic algorithm optimization on peptide sequences
• Screening peptide libraries with custom fitness functions (ML models, physicochemical filters)
• Generating, mutating, and recombining SMILES-based peptide representations
• Building multi-round directed evolution simulations in silico
• Integrating ML activity predictors into a peptide optimization loop

Installation

pip install pepflex==0.0.4
# Python ≥ 3.8

Core Classes at a Glance

Class	Role
PeptideGenerator	Generate random peptide sequences
Peptide	Single peptide with sequence, metadata, properties
PeptidePoolManager	Population container (add, retrieve, size)
PeptideMutator	Register and apply mutation rules
Evaluator	Pipeline of fitness functions + ranker
PoolRoundProcessor	Orchestrate one full evolution round

---

Quick Start — Full Evolutionary Loop

from pepflex import (
    PeptideGenerator, Peptide, PeptidePoolManager,
    PeptideMutator, Evaluator, PoolRoundProcessor
)
import pandas as pd

# 1. Generate initial pool
gen = PeptideGenerator()
initial_smiles = gen.generate_random_peptides(num_peptides=50, min_length=5, max_length=15)

pool = PeptidePoolManager()
for i, smiles_list in enumerate(initial_smiles):
    pool.add_peptide(Peptide(smiles_list, peptide_id=f"pep_{i}"))

print(f"Initial pool: {pool.get_pool_size()} peptides")

# 2. Configure mutations
mutator = PeptideMutator()
mutator.add_mutation_rule(mutation_type='n_terminal_addition', probability=0.3)
mutator.add_mutation_rule(mutation_type='inter_mutation',       probability=0.5)

# 3. Define evaluation pipeline (DataFrame-based)
def add_length(df): df["length"] = df["sequence"].str.len(); return df
def filter_min7(df): return df[df["length"] >= 7]
def my_scorer(df):   df["score"] = df["length"] * 0.1; return df  # replace with ML model

pipeline = [add_length, my_scorer, filter_min7]
ranker   = lambda df: df.nlargest(20, "score")

evaluator = Evaluator(evaluation_pipeline=pipeline, ranker_function=ranker)

# 4. Set up round processor
rp = PoolRoundProcessor()
rp.set_generation_function(
    lambda n: [Peptide(s, source_generation_params={"type": "replenishment"})
               for s in gen.generate_random_peptides(n, 5, 15)]
)

rp.add_pipeline_step('mutation',     rp._execute_mutation_step,
                     name='Mutate',  mutator=mutator, probability_of_application=0.8)
rp.add_pipeline_step('crossover',    rp._execute_crossover_step,
                     name='Crossover', num_crossovers=10, crossover_probability_per_pair=0.7)
rp.add_pipeline_step('evaluation',   rp._execute_evaluation_step,
                     name='Evaluate', evaluator_instance=evaluator)
rp.add_pipeline_step('replenishment',rp._execute_replenishment_step,
                     name='Replenish', target_size=50)
rp.add_pipeline_step('truncation',   rp._execute_truncation_step,
                     name='Truncate',  max_size=50)

# 5. Run N rounds
all_logs = pd.DataFrame()
for i in range(5):
    pool, logs = rp.run_round(pool, round_name=f"Round_{i+1}")
    all_logs = pd.concat([all_logs, logs], ignore_index=True)
    print(f"Round {i+1} — pool size: {pool.get_pool_size()}")

# 6. Inspect final population
top = pool.get_all_peptides()[:10]
for p in top:
    print(f"  {p.peptide_id[:8]}  1L={p.one_letter_sequence}  len={p.length}")