renunney

Working repository for replication and reanalysis of Leonard Nunney's 2003 cost-of-substitution simulations.

Biological Question

Nunney's paper asks how rapidly an adapting population can track a moving environment before extinction becomes likely. In this model:

• the selective optimum moves steadily through time,
• adaptation requires repeated allelic substitutions across one or more loci,
• population growth is density-dependent,
• offspring survival falls as genotypes lag behind the moving optimum,
• and "cost of substitution" is summarized by the smallest environmental-change interval T that still allows persistence.

Smaller T means faster environmental change and a harder adaptive problem.

Two Approaches

This project treats the problem in two separate ways.

Track 1: Nunney-faithful replication

• reconstruct the published simulation and threshold heuristic as closely as possible,
• preserve historically relevant assumptions even when they are inefficient or awkward,
• and use that result as the baseline for replication and criticism.

Track 2: modern replacement

• build a cleaner and faster simulator around the same biological question,
• define threshold explicitly rather than through Nunney's heuristic,
• and use a more performant implementation path, likely including Rust.

The current repo is centered on Track 1, with orchestration intended to support both tracks later.

Nunney's Approach

The paper-level model combines four pieces:

• constant environmental change,
• genotype survival w_i as a function of distance from the moving optimum,
• density-dependent female fecundity f,
• and Mendelian transmission with mutation.

The published threshold procedure is heuristic rather than inferential: Nunney searches for the smallest T with no extinctions in 20 runs, then checks nearby larger values. That historical rule is preserved in Track 1 because it is part of the claimed result structure.

This Repo's Approach

renunney turns that work into a clean, testable stack:

• local Track 1 simulation kernel,
• local threshold/search layer,
• local analysis, reporting, extinction-dataset, and fitting layers,
• local orchestration CLI and SQLite job registry,
• local paper-scale Figure 1 configs,
• and a Makefile for common operational tasks.

The repo was bootstrapped from earlier work in ../collaborations/to_ptbc/evc/cost_of_substitution, which remains useful as provenance and historical context, but the Track 1 runtime now lives in renunney.

Key Docs

• docs/MIGRATION.md
• docs/WORKFLOW.md
• docs/NUNNEY_ANALYSIS.md

Layout

• docs/ project, migration, and paper-analysis notes
• config/ configuration templates and paper-scale treatment configs
• runs/state/ SQLite registries and persistent orchestration state
• runs/results/ result artifacts collected by orchestration
• runs/scratch/ local worker scratch and cache files
• src/renunney/ in-repo Python package
• scripts/ local CLI entrypoints
• tests/ local verification

Start

Initialize the local run directories and SQLite registry:

make init

Run one local Track 1 simulation:

make track1-sim-smoke

Submit a paper-scale Figure 1 treatment:

make submit-figure1-m10

Run one worker loop locally:

make run-loop

Collate completed Figure 1 jobs:

make collate-figure1

Status

The Track 1 runtime and orchestration stack are now local to renunney. The next major step is no longer migration of Track 1 code; it is either:

• hardening multi-host orchestration,
• organizing publication-quality replication outputs,
• or starting the Rust-backed Track 2 path.