Navigability of genotype-phenotype (GP) maps

This code accompanies our Nature Ecology & Evolution paper and provides the computational tools for investigating properties of genotype-phenotype (GP) maps, specifically:

• Estimating GP map properties such as robustness and evolvability
• Estimate navigability of GP maps
• Simulate evolutionary dynamics

Install

To install and compile the required tools, packages and binaries, follow the instructions in INSTALL.md.

GP maps

GP maps on disk

Utilising GP map models anlaysed and published in 10.1371/journal.pcbi.1004773, we generate the following small GP maps and store on disk to facilitate faster computation:

• S_{2,8}: Polyomino GP map with two tiles and eight colours
• RNA12: RNA GP map with sequence length L=12
• HP5x5: Compact HP GP map on a 5x5 grid (L=25)
• HP3x3x3: Compact HP GP map on a 3x3x3 grid (L=27)
• HP20: Non-compact HP GP map of length L=20
• HP25: Non-compact HP GP map of length L=25

The GP maps are available after extracting:

tar -zxf gp_maps/gp_maps.tar.gz -C gp_maps
cd gp_maps

The GP maps have two files:

• geno_list*txt: line n is the phenotype of the the genotype that integer n is when represented in base K of the system.
• pheno_list*txt: the integer n in geno_list*.txt corresponds to the phenotype that is written on nth line OR for polyomino the file pheno*n*.txt which contains a diagrammatic representation of the polyomino.

The following numerical maps are used to convert character sequences to base K integer sequences:

- S_{2,8} : {'0': 0, '1': 1, '2': 2, '3': 3, '4': 4, '5': 5, '6': 6, '7': 7}
- RNA12   : {'A': 0, 'U': 1, 'C': 2, 'G': 3}
- HP5x5   : {'H': 0, 'P': 1}
- HP3x3x3 : {'H': 0, 'P': 1}
- HP20    : {'H': 1, 'P': 0}
- HP25    : {'H': 1, 'P': 0}
- fRNA    : {'C': 0, 'G': 1, 'A': 2, 'U': 3}

For example, the phenotype on line 12557964 of gp_maps/RNA_12/geno_list0.txt is phenptype index 56. This corresponds to phenotype on line 57 of gp_maps/RNA_12/pheno_list0.txt which is ((.((...)))). The mapping can be expressed as:

line of geno file  =      12557964   ->  genotype index     =      12557963
genotype index     =      12557963   ->  base K=4 (reverse) = '320223123332'
base K=4 (reverse) = '320223123332'  ->  RNA sequence       = 'GCACCGUCGGGC'
RNA sequence       = 'GCACCGUCGGGC'  ->  RNA dot-bracket    = '((.((...))))'

The above mapping can be applied to all GP maps on disk in the same fashion using the corresponding char to int maps described above.

GP maps constructed algorithmically

The S_{3,8}, RNA15 and fRNA GP maps are larger and the phenotype of a given genotype is algorithmically estimated as it is not computationally feasible to work with these larger GP maps on disk.

fRNA extraction

The functional RNA database (fRNAdb) is available to download.

We extracted all genotype sequences of length L with 20 <= L <= 140 from the file frnadb_summary.csv, outputting each sequence of a given <LENGTH> to fRNA/data/cs_l<LENGTH>.txt. Each sequence in fRNA/data/cs_l<LENGTH>.txt was then folded using ViennaRNA 1.8.5, with the dot-bracket structures saved in the same line by line order to fRNA/data/ps_l<LENGTH>.txt.

Experiments

GP maps stats

The properties of the GP maps on disk can be measured with the binary:

./bazel-bin/src/properties/gp_map_stats --help

For example, with RNA12 and component level estimates:

./bazel-bin/src/properties/gp_map_stats \
  --geno_fname gp_maps/RNA_12/geno_list0.txt \
  --outpath scratch/ \
  --base 4 \
  --length 12 \
  --nd_ref 0 \
  --nd_include 0 \
  --nd_threshold 1 \
  --estimate_component_stats \
  --verbose

To measure properties given random genotype swaps (e.g. 100,000):

./bazel-bin/src/properties/gp_map_stats \
  --geno_fname gp_maps/RNA_12/geno_list0.txt \
  --outpath scratch/ \
  --base 4 \
  --length 12 \
  --nd_ref 0 \
  --swaps 100000 \
  --seed 1 \
  --verbose

scripts/run_gp_map_stats.sh script to run the experiments for Table 1 and Fig. 2.

Navigability for GP maps on disk

Navigability can be estimated with any GP map in gp_maps/ with:

./bazel-bin/src/landscape/nav_disk --help

For example, with RNA12:

./bazel-bin/src/landscape/nav_disk \
  --geno_fname gp_maps/RNA_12/geno_list0.txt \
  --outpath scratch/ \
  --base 4 \
  --length 12 \
  --targets 50 \
  --sources 20 \
  --nd_ref 0 \
  --seed 1 \
  --dimension 4 \
  --swaps 100000 \
  --verbose

Navigability for GP maps not on disk

We estimate the navigability for larger GP maps that are not stored on disk by algorithmically folding/assembling. These are available for S_{3,8} and RNA15 here:

./bazel-bin/src/landscape/nav_alg --help

E.g. for S_{3,8}:

./bazel-bin/src/landscape/nav_alg \
    --base 8 \
    --length 12 \
    --file_name s38 \
    --nd_ref 1 \
    --neutral_mutations \
    --gp_map polyomino \
    --pheno_fname gp_maps/s_3_8/phenotypes/ \
    --assembly_tests 20 \
    --seed 1 \
    --sources $sources \
    --targets $targets \
    --threshold $threshold \
    --outpath scratch/

Navigability for restricted neutral correlations and dimensionality

scripts/run_nav_swaps_and_dim.sh provides a bash script to perform the experiments exploring how navigability changes under removing neutral correlations and restricting dimensionality is provided.

Navigability in fRNA

A binary for performing the navigability estimation in fRNA is available running:

./bazel-bin/src/landscape/nav_rna_dfs --help

Complete bash scripts for performing all fRNA experiments are available at scripts/run_nav_rna_dfs.sh. For example,

./scripts/run_nav_rna_dfs.sh 20000 hamming outs

will perform navigability estimation on fRNA at all lengths with and without neutral mutations with a threshold of 20000 total genotypes considered with Hamming fitness function applied.

Evolutionary navigability

Monormorphic evolutionary dynamics

A binary to run monormorphic evolutionary dynamics on RNA is available with:

./bazel-bin/src/evolution/nav_rna_mono --help

An example run is given below:

./bazel-bin/src/evolution/nav_rna_mono \
	--length 40 \
	--outpath scratch/mono/ \
	--fitness_measure hamming \
	--threshold 10000 \
	--sources 1 \
	--targets 1 \
	--pheno_fname "frna/data/ps_l40.txt" \
	--population_size 10000 \
	--norecord_diffs_only \
	--seed 1 \
	--neutral_mutations \
	--shape_level 0

A complete script for running the experiments is available at scripts/run_all_mono.sh which utilises a python subscript scripts/run_evolution_mono.py.

Polymorphic evolutionary dynamics

A binary to run monormorphic evolutionary dynamics on RNA is available with:

./bazel-bin/src/evolution/nav_rna --help

An example run is given below:

./bazel-bin/src/evolution/nav_rna \
	--length 40 \
	--outpath scratch/poly/ \
	--generations 10000 \
	--sources 1 \
	--targets 1 \
	--pheno_fname "frna/data/ps_l40.txt" \
	--population_size 100 \
	--seed 1

A complete script for running the experiments is available at scripts/run_all_poly.sh which utilises a python subscript scripts/run_evolution_polymorphic.py.

Analysis

The outputs of the evolutionary dynamics are in the form of sequential mutants (monomorphic case) or population summary statistics (polymorphic case). To derive estimates of the navigability additional analysis must be run on the output files to measure navigability in each source-target pair.

Scripts to produce these outputs across batches of paths containing evolutionary ouputs can be run with, for example:

python scripts/evo_analysis/run_analysis_script_batch.py\
       --paths \
	   data/raw_output/mono \
	   data/raw_output/poly \
       --executable scripts/evo_analysis/run_analysis_script.py \
       --by all

calling the subscripts scripts/evo_analysis/run_analysis_script.py and scripts/evo_analysis/analysis_script.py. with the outputs written within each of --paths to subpath analysis/.

Plotting

The plotting notebooks and scripts used to generate the figures and tables are available in plotting.