Nicholas W. VanKuren (1), Nathan P. Buerkle (1), Wei Lu, Erica L. Westerman, Alexandria K. Im, Darli Massardo, Laura Southcott, Stephanie E. Palmer, Marcus R. Kronforst

(1) Contributed equally to this manuscript. Contact Nick (nwvankuren@gmail.com; nvankuren@uchicago.edu) or Nathan (nathan.p.buerkle@gmail.com; nathan.buerkle@yale.edu) with questions about the supplied datasets or analysis.

Access this dataset on Dryad

Abstract

This Dryad repository contains raw data, scripts, and notebooks that can recapitulate the key analyses performed in the associated publication and allow further exploration the results. This repository contains three main sections:

1. Data, scripts, and a notebook detailing the genome-wide association analyses for forewing color and male mate choice. Data includes:
   • Raw forewing color and courtship behavior data from Chamberlain et al. (2009).
   • Genome-wide variant calls using whole-genome re-sequencing data from 113 H. cydno alithea males.
   • Raw and intermediate analysis results for color and choice genome-wide association tests
   • Intermediate results for analyses of linkage disequilibrium and $F_{ST}$ statistics
   • The H. c. alithea (yellow) reference genome sequence and gene annotation generated and used in this paper
2. Data, scripts, and a notebook detailing the analysis of RNA-sequencing data for H. c. alithea and H. c. galanthus central brain, optic lobe, and retina at seven developmental stages. Data includes:
   • Intermediate gene expression data
   • Full results from differential expression analyses presented the paper
3. Raw and processed electrophysiology data for individual photoreceptors, labeled by species, sex, and individual; scripts for processing those data and reproducing publication figures. This directory was updated on 2026-08-31 to include raw ABF files and additional processing scripts. Check the "Versions" section below and the README files contained within the electrophysiology.tar.gz directory for more information. Data includes:
   • Raw ABF files from single-cell recordings
   • Tuning curve response amplitudes and latencies
   • V-Log(I) response amplitudes and latencies
   • Tuning curves in the presence of green LED

Versions

• 2026-08-31 Update:

  Updated to include raw and processed data and a reproduction pipeline for the electrophysiology figures (Figures 4, 5 and S7 to S10) of VanKuren & Buerkle et al. (2025) PLOS Biology 23(3): e3002989. Dates are ISO (YYYY-MM-DD).

  The original Dryad deposit consisted of six summary tables (UV/blue/green/red/broadband_photoreceptors.csv and wing_reflectance_data.csv) holding per-cell metadata, tuning curves, latencies, V-log(I) values and LED-condition values, with no voltage traces and no code for analyzing them. This deposit is built from the same 508 recordings (UV 180, blue 126, green 149, red 33, broadband 20) and adds the recordings themselves, the code that produces every figure, and per-cell tables that were not previously available.

  • Revised Methods/Electrophysiology methods section (above) to accurately reflect our data collection and analysis. These changes are also detailed in the associated updates in the revised electrophysiology.tar.gz directory and below.

  • Added raw and processed data files

    Additional details on files can be found in the "Description of the data and file structure" section below.

    • Per-photoreceptor-type MATLAB v7.3 files {uv,blue,green,red,broadband}_photoreceptors.mat, one self-contained record per cell holding metadata (species, sex, lambda_max, inhibition label), tuning curves, the V-log(I) intensity series where one was recorded, and, for all types except broadband, the voltage traces. Traces are 6000 samples at 10 kHz. Traces from 2019 onward include a second channel with the shutter TTL and are aligned so the TTL edge lies at sample 2000; earlier traces are single channel and unaligned. A consolidated led_data.mat is included.
    • Python code to regenerate every figure from the data, driven by generate_all_figures.py, together with load_published_data.py, which provides the loaders and a response-latency routine that times the depolarizing and hyperpolarizing components separately.
    • Both kinds of tuning curve, reported separately. The paper uses "sensitivity" for two different quantities: the normalized voltage curve, and the curve obtained by inverting the Naka-Rushton fit. Each tuning_curve now reports both, under names that tell them apart. dominant_mV is the response in millivolts and response_normalized is that divided by its peak; both are signed, so long wavelength inhibition appears as a negative tail. sensitivity is the transformed curve and sensitivity_normalized is that divided by its peak; the transform is undefined for responses at or below zero and floors there, so neither of those two can represent a hyperpolarizing response. The original Dryad deposit published the transformed curve where a V-log(I) series was available and the normalized voltage curve for cells without one, both under a single heading. The transform requires a V-log(I) series. Of the 508 cells, 323 have one and 185 do not. Six of the 323 gave no converged Naka-Rushton fit, leaving 317 cells with a transformed curve, of which 129, 41 percent, contain at least one wavelength at which the response is floored. The sensitivity fields are set to NaN for the other 191 cells, the 185 with no series plus those six.
    • The ABF link mirrored into each cell record as abf_session, abf_cell_number and abf_match_correlation, so a single MAT file answers where a cell came from. Present for all 488 cells that have traces, including the 65 from the 2017 sessions whose numbering was recovered by waveform matching, and blank for the 20 broadband cells.
    • abf_cell_map.csv, linking each cell to the raw recording it came from, by session and cell number. The mapping was established by matching the deposited traces against the original recordings and is exact for all 488 cells that have traces. Broadband cells have no such linkage.
    • vlogi_series.csv, intensity-response series for the 222 cells whose series was recovered from the session index files: UV 30, blue 85, green 91 and red 16. The original Dryad deposit included these for 151 cells. A V-log(I) series is held for 323 cells in all; the 101 UV cells not in this file are those whose series comes from the published tables rather than from the session files, and they are reachable through the MAT records. Those records embed every series as cell.vlogi, holding the intensities, the responses and a Naka-Rushton fit, for all 323 cells: UV 131, blue 85, green 91 and red 16. Broadband has no intensity series. The Naka-Rushton parameters stored with each series come from two sources, told apart by the sign of n. For 101 UV cells the values are those fitted for the original Dryad deposit, which reproduce its sensitivity curves. For the other 222 series, which were recovered from the session indices and have no published parameters, the values are a fit made here of V = Vmax*I^n/(I^n + k^n) with I = 10^(-intensity_log). Six fits that did not converge are set to NaN.
    • glme_results.csv, the fixed-effect coefficients of the generalized linear mixed-effects models the paper reports for Figures 4B, 5I and 5J, with butterfly identity as a random effect. The reproduction scripts for those three figures display the mixed-model value, which is the statistic the paper reports. On Figure 4B the white against yellow H. c. alithea bracket shows the mixed-model value, p = 0.98, which is the figure the paper quotes. On Figure 5I the asterisks mark the pairwise comparisons the caption names, at p < 0.01: white against yellow H. c. alithea, H. c. galanthus against H. pachinus, and H. c. galanthus against the F1 hybrids. ERRATA.txt item 11 maps each model to the claim it supports and sets out the scope of the Methods sentence describing the binary alternative to the response-amplitude measure. The published p-values are reproduced exactly by the deposited code.
    • lambda_max_fits.csv, rhodopsin-template fits with diagnostics, including the Naka-Rushton parameters used to convert responses to spectral sensitivity and an adopted column marking which of the fits are the values stored in the MAT files.
    • wavelength_presentation_orders.csv and companion files giving the order in which stimuli were presented.
    • ERRATA.txt, listing points at which the published Methods text does not fully and clearly describe what was done.

  • Explicitly calculated values that differed from the original Dryad deposit

    Inhibition labels, sex, latencies and the LED-condition values are unchanged. Two quantities differ, in both cases deliberately.

    • Tuning curves. The original Dryad deposit published one curve per cell: the transformed curve where a V-log(I) series was available and the normalized voltage curve otherwise, both under a single heading. The transform is undefined for responses at or below zero and floors there, so a transformed curve cannot represent a hyperpolarizing response. In the earlier deposit this left 53 of the 76 UV cells labeled as inhibited with curves that never fall below -0.001. This deposit reports the response voltages in millivolts alongside the normalized curves, so that the long wavelength inhibition described in the paper is directly measurable. The same 76 cells reach a median of -0.196 normalized units at 530 nm in the response voltages. Group proportions and every published statistic are unchanged.

    • lambda_max. Values for UV, blue and green cells were refit by converting each cell's responses to spectral sensitivity using its own Naka-Rushton fit and then fitting the standard rhodopsin template, which is the procedure the paper describes. Fitted values track the original Dryad deposit closely. Quartiles, in nm:

      type	n	regenerated	original Dryad deposit
      UV	179	361.67 / 373.26 / 379.49	361.15 / 371.61 / 378.13
      blue	126	448.09 / 451.16 / 453.89	447.54 / 451.61 / 455.22
      green	149	543.77 / 551.18 / 557.36	541.49 / 548.62 / 555.74

      One UV cell has no intensity series, yielded no fit, and is not counted above. Red and broadband are set to NaN, matching the N/A recorded for every red and every broadband cell in the original Dryad deposit. One UV cell, index 148 shows a hyperpolarizing offset that exceeds its depolarization at every wavelength; its tuning curve holds the depolarizing component, as described below. Fits are regenerated by code/fit_lambda_max.py.

      Figure 4B is the one figure that does not use these refits. It reads lambda_max from data/dryad_csvs/UV_photoreceptors.csv so that the panel reproduces the published values rather than the refit. The refit runs between 1.11 and 2.03 nm above the published value on the six groups of males, and the cell counts are identical, 43, 40, 18, 19, 8 and 30 for those groups and 22 females, matching the published caption.

  • Added an explanation for the tuning curve of UV cell 148

    dominant_mV holds, per wavelength, whichever of the depolarizing and hyperpolarizing components has the larger magnitude, with its sign preserved. UV cell 148 shows a hyperpolarizing offset of roughly constant amplitude, between -9.6 and -11.9 mV, at every one of the 38 wavelengths, which exceeds its depolarization across the whole spectrum. For that cell the tuning curve holds the depolarizing component instead, which is spectrally tuned and peaks at 9.26 mV at 340 nm. depol_mV and hyperpol_mV are unchanged, so the larger-magnitude value can be recomputed per wavelength by anyone who wants it. This is the only cell in the deposit that meets the condition. The sensitivity and response_normalized curves for that cell follow from the same component.

  • Added functions to reproduce Figure 5A

    The two example cells are those in the published panel: cell 10, trial 0, from session aliWM_03-Oct-2019, and cell 39, trial 0, from aliWM_07-Oct-2019, both white H. c. alithea males. The published panel labels them UV Cell 1 and UV Cell 2.

  • Added functions to reproduce Figure 4A

    • The dotted reference curves are read from data/fig4A_published_templates.csv, which holds the UV1 and UV2 curves as they are drawn in the published panel, from 310 to 450 nm at 1 nm, each normalized to a peak of 1. The values were measured from the published figure.
    • The measured curves are the mean and SEM of the tuning curve column of data/dryad_csvs/UV_photoreceptors.csv, over the same 83 H. c. alithea males and 22 females as the published caption. That column holds a mixture of V-log(I) transformed sensitivity and normalized voltage across cells, as described above. The response voltages held in the MAT files are shown in Figure S7.

  • Added functions to reproduce statistical tests in the figures

    Each figure script computes the test its published caption names, and no other.

    • Figure 4B. Drawn from the published lambda_max values. Asterisks below mark a significant difference from the expected tuning of both UV1 and UV2, by one-sample t-test against each template with Bonferroni correction. The white against yellow H. c. alithea bracket shows the pairwise mixed-model value.
    • Figure 5H. Asterisks above mark a significant change from before the LED, by one-sample t-test of each group against zero. The between-group comparison is a two-sample t-test. This reproduces the values in the text: -5.6 +/- 1.1 mV with p < 0.001 for cells with hyperpolarizing responses, and -1.4 +/- 0.8 mV with p = 0.10 for those without.
    • Figures 5I and 5J. Asterisks mark pairwise mixed-model differences between groups at p < 0.01.

  • Clarified conventions

    • Attenuation is given in log units, 0 being the brightest.
    • Latencies are relative to light onset, taken as the TTL edge plus an 11.9 ms mechanical shutter delay, and can be measured only for cells recorded with a TTL channel.
    • Every data file has a companion section in data/README.md.

• 2024-12-09 Update:

  • Updated title to reflect final title.
  • Added a new directory, electrophysiology.zip, containing raw data from electrophysiology experiments and wing spectral reflectance data. Updated README to describe new files.
  • Updated Methods section to describe electrophysiology data collection

• 2024-10-18 Update:

  • Updated title to reflect order of paper results.
  • Updated Methods section of this repository to add information of how data required for analyses described in the gwas directory were generated.
  • Minor updates to initial rnaseq submission, including: minor text edits in notebook.Rmd for clarity; the addition of an updated file data/GenesInOtherPrefPeaks.50kb.2024-10-10.txt containing a list of genes found near genome-wide association (GWA) peaks on chromosomes 7, 9, and 11.
  • Relocated top-level genome and genome annotation files into gwas/info/genome directory.
  • Added a new gwas directory containing data, scripts, and an R notebook required for running the GWA analyses presented in the associated publication, plus additional QC steps that were not presented in the publication but are important for understanding our final analyses. This README has been updated to reflect this addition and describes each of those files below. The gwas/notebook.Rmd R notebook contains additional information and will knit properly after installation of the required R packages.

Description of the data and file structure

Description of data and files within the gwas directory:

The most important files in this directory are the notebook.Rmd and the gwas_notebook.Rproj files. These two files can together be used to re-run analyses presented in the paper, contain much of the detail and reasoning behind each analysis, and link all of the data and scripts described below into a single cohesive story. Please start there.

gwas.tar.gz

• gwas_notebook.Rproj: R project file for loading into R and RStudio. In conjunction with notebook.Rmd, can be used to re-run analyses presented in the associated paper and to run new analyses of interest to end user.
• notebook.Rmd: R markdown file containing detailed notes, code chunks, and additional data processing steps that can be used to recapitulate the paper results and allow exploration of the results further.
• expected_notebook.html: output from knitting notebook.Rmd before dryad submission. End users may simply browse this notebook or knit the project fresh.
• cluster_scripts/
  • PreprocessReads.sh: a shell script for SLURM that can be used to perform quality control and mapping of whole-genome re-sequencing data for the samples found in BioProject PRJNA802836
  • RunHaplotypeCaller.sh: a shell script SLURM that can be used to perform the first step of calling genome-wide variation in the alignment files generated by PreprocessReads.sh. This script depends on the GenomeAnalysisToolKit v4.2 and the HaplotypeCaller module.
  • RunGenotypeGvcfs.sh: a shell script for SLURM that can be used to perform the second step of calling genome-wide variation using the output from RunHaplotypeCaller.sh. This script depends on the GenomeAnalysisToolKit v4.2.
  • RunGemmaFwcGwa.sh: a shell script for SLURM that can be used to perform the genome-wide association for male forewing color (fwc) using the variant callset found in data/plink/heliconius_cydno_alithea_yellow.variants.20230712.pruned...
  • RunGemmaLmpGwa.sh: a shell script for SLURM that can be used to perform the genome-wide association for male lifetime mate preference (lmp) using the variant callset found in data/plink/heliconius_cydno_alithea_yellow.variants.20230712.pruned...
  • RunGmmatChoiceWald.R: an R script that will perform the genome-wide association for choice using GMMAT 1.4.2, the variant callset in data/plink/heliconius_cydno_alithea_yellow.variants.20230712.pruned... and the raw phenotype and choice data in data/phenotype_data/MalePhenotypeData.20190306.txt and data/phenotype_data/CourtsForR.20190506.EXPANDED.csv.
  • RunGmmatChoiceWald.sh: a shell script for SLURM that wraps RunGmmatChoiceWald.R.
  • RunGmmatFwcWald.R: an R script that will perform the genome-wide association for male forewing color (fwc) using GMMAT 1.4.2, the variant callset in data/plink/heliconius_cydno_alithea_yellow.variants.20230712.pruned... and the raw phenotype data in data/phenotype_data/MalePhenotypeData.20190306.txt.
  • RunGmmatFwcWald.sh: a shell script for SLURM that wraps RunGmmatFwcWald.R.
• current_cache/: an empty directory to contain cached results from knitting the R notebook ./notebook.Rmd
• data/
  • fst/
    • heliconius_cydno_alithea_yellow.variants.20230712.pruned.10k2k.windowed.weir.fst.gz: gzipped text file containing $F_{ST}$ calculations in 10 kb sliding windows (2 kb step) between yellow and white males. Calculated using vcftools 0.16 and the variant calls in data/plink/heliconius_cydno_alithea_yellow.variants.20230712.pruned...
  • gwas/
    • choice.grm_only.20240923.out.txt.gz: gzipped text file containing the genome-wide association results for male mate choice including only the genetic relatedness matrix (grm) as a random effect.
    • choice.grm_stock.20240925.out.txt.gz: gzipped text file containing the genome-wide association results for male mate choice including the genetic relatedness matrix (grm) and principal component 2 (pc2, or stock) as random effects.
    • forewing_color.cXX.txt: the genetic relatedness matrix calculated using GEMMA 0.98.5 and the variant callset in data/plink/heliconius_cydno_alithea_yellow.variants.20230712.pruned...
    • forewing_color.grm_only.assoc.txt.gz: gzipped text file containing the genome-wide association results for male forewing color including only the genetic relatedness matrix as a random effect.
    • forewing_color.grm_pc2.assoc.txt.gz: gzipped text file containing the genome-wide association results for male forewing color including the genetic relatedness matrix and principal component 2 (pc2, or stock) as random effects.
    • forewing_color.grm_pcs1-3.assoc.txt.gz: gzipped text file containing the genome-wide association results for male forewing color including the genetic relatedness matrix (grm) and principal components 1-3 as random effects.
    • klocus.court_num.20240926.out.txt.gz: gzipped text file containing the association results for male mate choice in the K locus (chr1:14.5Mb-16.5Mb) including the genetic relatedness matrix (grm) a random effect and court number as an additional fixed effect.
    • klocus.male_fwc.20240926.out.txt.gz: gzipped text file containing the association results for male mate choice in the K locus (chr1:14.5Mb-16.5Mb) including the genetic relatedness matrix (grm) a random effect and male forewing color as an additional fixed effect.
    • preference_all.grm_only.assoc.txt.gz: gzipped text file containing the genome-wide association results for male lifetime preference (estimated using all courts) and including only the genetic relatedness matrix as a random effect.
  • ld/
    • K_locus_LD.lt10kb.ld.gz: gzipped text file containing pairwise linkage disequilibrium (LD) estimates for all pairs of K locus variants less than 10 kb apart.
    • LdEstimates.100Mb.240905.txt: text file containing summaries of LD values for 600 million random pairs of variants from across the genome.
    • snp_ld_values: text file containing pairwise LD estimates for pairs of top forewing color and male mate choice variants identified using genome-wide association.
  • plink/
    • heliconius_cydno_alithea_yellow.variants.20230712.pruned.[bed|bim|fam|nosex]: plink bfile set containing filtered variant calls for all 113 males in the final dataset. This was the callset used for all genome-wide association analyses.
    • heliconius_cydno_alithea_yellow.variants.20230712.pruned.eigenv[ec|al]: text files containing principal components (PC) analysis results for the first 20 PCs. Calculated using PLINK 1.90 and the plink bfile set in this directory.
    • heliconius_cydno_alithea_yellow.variants.20230712.pruned.forewing_color.fam: plink fam file containing the male forewing color information.
  • susie/
    • SusieData.20241003.Rdata: R data file containing two objects used for the SuSiE analysis described in the publication and the notebook.
• info/
  • genome/
    • heliconius_cydno_alithea.yellow.fa.gz: bgzipped FASTA format file containing the yellow Heliconius cydno alithea genome sequence.
    • heliconius_cydno_alithea.yellow.agp: text file containing information linking heliconius_cydno_alithea.yellow.fa scaffolds to the Heliconius melpomene v2.5 genome assembly. Generated using RagTag and subsequent massaging with awk.
    • heliconius_cydno_alithea.yellow.genome: text file containing heliconius_cydno_alithea.yellow.fa scaffold names and lengths.
    • heliconius_cydno_alithea.yellow.decorated.modified.gff3.gz: gzipped generic feature format v3 (GFF3) file containing gene annotation information for the heliconius_cydno_alithea.yellow.fa genome sequence.
  • phenotype_data/
    • CourtsForR.20190506.csv: csv file containing raw courtship data from Chamberlain et al. (2009)
    • CourtsForR.20190506.EXPANDED.csv: csv file containing raw courtship data from Chamberlain et al. (2009) EXPANDED to include a single courtship event per row.
    • GenotypePhenotypeCourtshipDataForR.20190620.txt: tab-delimited file containing genotype and courtship data for males studied by Chamberlain et al. (2009)
    • MalePhenotypeData.20190306.txt: tab-delimited file containing male wing color pattern information from Chamberlain et al. (2009)
  • sequencing_qc/
    • Raw-Alithea-Data_multiqc_report[.html|_data.zip]: html file and associated data summarizing the quality of the raw whole-genome re-sequencing data using FastQC and MultiQC. Data can be downloaded from NCBI BioProject PRJNA802836.
    • Trimmed-Alithea-Data_multiqc_report[.html|_data.zip]: html file and associated data summarizing the quality of the processed whole-genome re-sequencing data using FastQC and MultiQC. Data can be downloaded from NCBI BioProject PRJNA802836 and processed using cluster_scripts/PreprocesReads.sh.
    • SequencingQcAndMappingForR.20190620.txt: text file containing summary data for re-sequenced samples.
  • plots/: empty directory for holding plots generated by knitting the notebook.Rmd file in RStudio.
• scripts/
  • RunGmmatScore.R: R script to run the genome-wide assocation analysis for male mate choice using GMMAT and the score test.
  • functions_for_gwa.R: R script containing useful functions for loading data, plotting, and massaging data.

Description of data and files within the rnaseq directory:

The most important files in this directory are the notebook.Rmd and the rnaseq.Rproj files. These two files can together be used to re-run analyses presented in the paper, contain much of the detail and reasoning behind each analysis, and link all of the data and scripts described below into a single cohesive story. Please start there.

rnaseq.tar.gz

• rnaseq.Rproj: R project file for loading into RStudio
• notebook.Rmd: R markdown notebook containing detailed notes, pipelines, and code chunks for recapitulating the published results and to allow further independent exploration by the end user.
• data/
  • AllGeneFunctionalAnnotations.2023-04-04.csv.gz:
  • GenesInKlocus.2023-05-23.txt: A text file containing a list of genes within the K locus (Hcay201001o:14500000-16500000)
  • GenesInOtherPrefPeaks.50kb.2024-10-10.txt: A text file containing a list of genes within 50 kb of the top GWA peaks on chromosomes 7, 9, and 11.
  • heliconius_cydno_alithea_deseq_results.Rdata: R data file containing the results from all stage-specific differential expression analyses used in the publication. This is a list of tibbles, deseq_results, where each tibble contains the full results from a single stage-specific run of DESeq2. This can be used to analyze stage-specific DE results for any gene.
  • helicoinius_cydno_alithea.yellow.eggnog.csv.gz: A gzipped CSV file containing the results from running eggNOG's emapper utility on the full yellow Heliconius cydno alithea protein annotation set. This contains information of putative orthologs and functions for each gene.
  • heliconius_cydno_data.Rdata: An Rdata file containing four objects used to run the publication analyses.
    • tx2gene: a 2-column tibble containing transcript-to-gene mapping
    • sample_info: a tibble containing RNA-seq sample information
    • initial_dds: a DESeq2 data object containing unfiltered gene quantification data for each sample in sample_info. Values were loaded from raw salmon output for each sample using tximport package.
    • filtered_dds: a DESeq2 data object containing filtered gene quantification data for the final set of 260 samples included in the publication analyses. This is the initial_dds object filtered for genes with mean normalized expression values >50 and for non-outlier samples.
  • heliconius_cydno_DEGs.ym_v_wm.gFDR_0.01.Rdata: An Rdata file containing a list of vectors of DE genes, DEGs_list, found in each relevant DESeq2 and maSigPro comparison. This is deseq_results filtered to keep only genes with global FDR <= 0.01 for each comparison, plus genes found to be DE by maSigPro. See the notebook for details on how these vectors were generated.
  • heliconius_cydno_masigpro_results.Rdata: An Rdata file containing the results from maSigPro runs from each tissue. This file contains list objects for each tissue (central brain, optic lobe, and retina) with the raw maSigPro results. You can extract significant genes using maSigPro::get.siggenes().
  • heliconius_cydno_wgcna_objects.Rdata: An Rdata file containing the objects required for running and generated by WGCNA with the filtered_dds and sample_info information. This file contains:
    • datExpr: expression data, generated from DESeq2::counts(filtered_dds, norm = T)
    • datTraits: trait data, derived from sample_info
    • dynamicMEs: gene co-expression module eigenvectors (summaries of gene expression for all genes within that module), for dynamically pruned modules.
    • dynamicModuleColors: vector containing module colors for each gene included in datExpr. These are the module assignments given to each gene by running WGCNA.
    • dynamicModuleLabels: vector containing module labels (numbers) for each gene included in datExpr. These are the module assignments given to each gene by running WGCNA.
    • gene_tree: hierarchical clustering object (from hclust) describing the relationships of all genes in datExpr. You can plot using base::plot().
    • METree: hierarchical clustering object (from hclust) describing the relationships of all co-expressed gene modules, based on the eigenvectors contained in dynamicMEs. You can visualize using base::plot
• plots/publication_plots/rPCA outlier identification/: directory containing the rPCA outlier plots that we relied on for the publication's filtering and analysis steps. These may differ from the analogous plots generated by the notebook because of random seeding. Files are labeled 9-OutlierId.all_{stage}_{tissue}.outliers.png. Where stage and tissue are described in the notebook.
• results/module_GOs/
  • go_enrichment/: directory containing topGO gene ontology (GO) enrichment results in CSV format for each co-expressed gene module identified in the publication analyses. Module colors match those in the dynamicModuleColors vector.
  • go_figure/: directory containing plots generated using GO-figure!, the results from WGCNA, and the gene annotations held in heliconius_cydno_alithea.yellow.eggnog.csv.gz. The results for each invocation are found in independent directories. The topgo_files directory contains the properly formatted input files, derived from the eggNOG annotations.
  • graphs/: empty directory to contain output from topGO runs produced during knitting.
• scripts/:
  • heliconius_RNAseq_functions.R: R script file containing custom functions that I used to produce plots, massage data, etc.
  • modified_masigpro_functions.R: R script file containing maSigPro functions that have been modified to produce more useable/aesthetically pleasing output. These supersede functions in the native maSigPro package when knitting this notebook.

Description of data and files within the electrophysiology directory:

electrophysiology.tar.gz

• CHANGELOG.md: Description of changes made between the 2024-12-09 and 2026-08-31 updates.
• ERRATA.txt: Points at which the published Methods text does not fully and clearly describe what was done, with the correct statement and how it was established in each case. None of them alters the results presented or the conclusions drawn in the published version of the paper.
• README.md: Overview of and directions for using the data and scripts contained within the electrophysiology directory.
• requirements.txt: List of required python packages for running scripts to recapitulate figures.
• code/: directory containing python scripts to access and process raw electrophysiology data and to generate publication figures. Instructions for using each script are found in the top level README.md file.
  • figure_vlogi_other_types.py: Python3 script to plot V-log(I) (intensity-response) for blue / green / red photoreceptors, recovered from the per-session index files (vlogiData) bundled with the raw ABFs. Companion to Figure S9 (UV).
  • figure4A_spectral_sensitivity.py: Python3 script to reproduce Fig4A displaying UV photoreceptor spectral sensitivity of H. c. alithea males vs all females.
  • figure4B_lambda_max_preference.py: Python3 script to reproduce Fig4B showing UV photoreceptor lambda_max by species.
  • figure5A_example_traces.py: Python3 script to reproduce Fig5A showing example photoreceptor voltage traces.
  • figure5B_UV_tuning.py: Python3 script to reproduce Fig5B showing mean tuning curves for UV cells with and without inhibition.
  • figure5C_UV_latency.py: Python3 script to reproduce Fig5C showing UV photoreceptor response latency vs wavelength, for cells WITH (n=76) vs WITHOUT (n=104) long-wavelength hyperpolarizing responses (the groups of Fig 5B).
  • figure5D_blue_tuning.py: Python3 script to reproduce Fig5D showing blue cell tuning curves.
  • figure5E_blue_latency.py: Python3 script to reproduce Fig5E showing blue cell latency.
  • figure5F_green_tuning_LED.py: Python3 script to reproduce Fig5F showing green cell tuning curves with and without green LED background.
  • figure5G_UV_tuning_LED.py: Python3 script to reproduce Fig5G showing UV photoreceptor tuning curves Before vs During the green LED background.
  • figure5H_UV_LED_lambda_max_Vrest.py: Python3 script to reproduce Fig5H showing the effects of green LED effects on UV cell max response and resting potential.
  • figure5I_UV_proportion_inhibited.py: Python3 script to reproduce Fig5I showing the proportion of UV photoreceptors with LW inhibition by species.
  • figure5J_blue_proportion_inhibited.py: Python3 script to reproduce Fig5J showing the proportion of blue photoreceptors with LW inhibition by species.
  • figureS10_temporal_responses.py: Python3 script to reproduce FigS10 showing photoreceptor temporal responses.
  • figureS7_photoreceptor_tuning.py: Python3 script to reproduce FigS7 showing photoreceptor sensitivities.
  • figureS8_lambda_max_distributions.py: Python3 script to reproduce FigS8 showing lambda max values for each photoreceptor type in each group.
  • figureS9_v_log_i_comprehensive.py: Python3 script to regenerate FigS9 displaying V-log(I) intensity-response and response latency vs intensity.
  • fit_lambda_max.py: Python3 script to regenerate data/lambda_max_fits.csv from the deposited data.
  • generate_all_figures.py: Python3 script to generate ALL reproduced electrophysiology figures.
  • load_published_data.py: Python3 script containing helper functions to load data from published MAT files.
• data/: directory containing raw and processed electrophysiology data files.
  • README.md: File containing information on directory contents.
  • abf/: directory containing raw recordings from each session, 90 tarballs in total. Each tarball holds the ABF files for every cell and wavelength recorded in that session.
    • manifest.csv: Parts list describing contents of each tarball
    • {session}.tar.gz: tarballs containing raw data, named by recording session, for example aliWM_03-Oct-2019.tar.gz. Extract using tar -xzf ...
  • abf_cell_map.csv: One row per cell giving abf_session and abf_cell_number, identifying the raw recording the traces came from.
  • dryad_csvs/: directory containing the six per-cell summary tables released with the 2024-12-09 version of this deposit, holding metadata, tuning curves, latencies, V-log(I) values and LED-condition values for each photoreceptor type. Retained so that the published values remain available alongside the rebuilt records.
  • fig4A_published_templates.csv: The UV1 and UV2 reference curves as drawn in the published Figure 4A, from 310 to 450 nm at 1 nm, each normalized to a peak of 1. Read by figure4A_spectral_sensitivity.py.
  • *.mat files: MATLAB struct arrays with one entry per cell, separated by cell type. These files can be read and analyzed using MATLAB and procedures described in the README.md file. led_data.mat holds the LED-condition measurements consolidated across cell types: tuning curves and V-log(I) series recorded before and during the green adapting LED, and the accompanying resting potentials.
  • lambda_max_fits.csv: Rhodopsin-template fits for the 455 UV, blue and green cells, with fit_r2, the source of the V-log(I) used, and the Naka-Rushton parameters.
  • glme_results.csv: Fixed-effect coefficients of the generalized linear mixed-effects models the paper reports for Figures 4B, 5I and 5J. Each model has the form response ~ 1 + effect + (1 | subject), with butterfly identity as the random effect so that several cells from one animal are not treated as independent, fitted in MATLAB with fitglme by maximum pseudo-likelihood and, for the binary inhibition classification, a binomial distribution with a logit link.
  • vlogi_series.csv: Intensity-response series, one row per cell and intensity level, with columns intensity_log (log attenuation, 0 brightest), response_mV, wavelength_nm and led.
  • uv_vlogi.csv / uv_vlogi_led.csv: UV V-log(I) intensity series, including inhibitory-wavelength responses and per-intensity latency. Used for Figure S9.
  • uv_led_resting_potential.csv / uv_led_tuning.csv: Files containing UV cell measurements in the presence of a green LED.
  • green_led_tuning_updated.csv: File containing green cell tuning in the presence and absence of LED.
  • vlogi_blue_green_red_broadband_README.md: README describing the contents and usage of the subsequent vlogi_blue_green_red_broadband.csv file.
  • vlogi_blue_green_red_broadband.csv: Blue, green and red V-log(I) series, for 107, 104 and 9 cells respectively, plus 14 further cells whose type assignment is uncertain and is marked as such. Used by figure_vlogi_other_types.py. There is no broadband V-log(I). See its _README.md.
  • vlogi_latency_blue_green.csv: Blue/Green V-log(I) latency-vs-intensity (Fig S9 Panel B).
  • wavelength_presentation_orders_README.md: README describing the contents and usage of the subsequent wavelength_*csv files.
  • wavelength_presentation_orders.csv: The seven fixed wavelength presentation orders used study-wide, Order_A to Order_G, one column per sequence (play_position by order). See its _README.md.
  • wavelength_order_by_session.csv: Orders of stimulus presentation used in each recording session.
  • wavelength_order_by_cell.csv: Orders of stimulus presentation for each cell.
  • led_data.mat, *_led_*.csv: LED-condition tuning / V-log(I) / resting-potential data.
• figures/: Directory containing the output from running electrophysiology/code/generate_all_figures.py. all_plots.pdf contains all plots in PDF format. Additional files in PNG or PDF format were generated using the matching scripts in the code/ directory.

Sharing/Access information

Raw whole genome sequencing data can be downloaded from the NCBI Short Read Archive through BioProject PRJNA802836.

Raw RNA-sequencing data can be downloaded from the NCBI SRA through BioProject PRJNA1019262.

Code/Software

R notebooks

All R work was performed in RStudio 2024.04.1+748 and R version 4.4.

R notebooks contained in the gwas and rnaseq directories contain all information needed to run and should install required packages upon first execution. The expected_notebook.html files contain the expected results from simply opening the notebooks in RStudio and knitting using knitr. Packages and version numbers are contained in those html files via sessionInfo().

All Rcode contained within the gwas and rnaseq directories should be self-contained and able to run on standard system setups. NWV executed R notebooks on a 2015 MacBook Pro with 16Gb RAM and 2.5GHz i7 processor running Monterey 12.7.5. W Lu executed on a Windows machine with 16Gb RAM, 2.4GHz i7 processor and running Windows 10.

R notebooks contain additional commands that we used to run standalone programs on a compute cluster or on the UNIX command line. These are contained within their own chunks within those notebooks. These chunks were used to produce many of the intermediate results files found in data, and the input and output to those commands is written relative to the respective directory.

Python scripts

All electrophysiology processing and figure code is Python 3 and is contained in the electrophysiology/code directory. Required packages are listed in electrophysiology/requirements.txt: numpy 1.20 or later, scipy 1.7, matplotlib 3.4, h5py 3.0 and pandas 1.3. These can be installed with pip install -r requirements.txt.

To regenerate every figure, run the following from the electrophysiology directory:

python3 code/generate_all_figures.py

This writes 17 figures to figures/, each as a 300 dpi PNG and a vector PDF, together with figures/all_plots.pdf holding all of them. Individual panels can be produced by running the corresponding script in code/ the same way.

The scripts read the deposited .mat files directly using h5py, since MATLAB v7.3 files are HDF5, so no MATLAB license is required to use the data or to reproduce the figures. The same files can also be opened in MATLAB directly. The generalized linear mixed-effects models the paper reports were fitted in MATLAB with fitglme; their coefficients are deposited in data/glme_results.csv and are read from that file rather than refitted.

SEP executed the electrophysiology scripts on a 2023 MacBook Pro with 16Gb RAM and an Apple M2 Pro processor running Ventura 13.5.1, using Python 3.14.3.

SLURM scripts

It is likely that users will need to edit the SLURM directives in any of the cluster_scripts scripts for their particular setups, in particular the account and the module loading steps.

Additional programs

• GenomeAnalysisToolKit 4.2
• PLINK 1.90
• GEMMA 0.98.5
• samtools, bcftools, and htslib 1.20
• bowtie2 2.5.1
• trim_galore
• cutadapt 4.5
• FastQC 0.12.1
• MultiQC 1.21
• picard 2.8.1
• salmon 1.10.1
• GOFigure! 1.0.2