This repository contains code used in our manuscript How useful is genomic data for predicting maladaptation to future climate?
Abstract
Methods using genomic information to forecast potential population maladaptation to climate change or new environments are becoming increasingly common, yet the lack of model validation poses serious hurdles toward their incorporation into management and policy. Here, we compare the validation of maladaptation estimates derived from two methods – Gradient Forests (GFoffset) and the Risk Of Non-Adaptedness (RONA) – using exome capture pool-seq data from 35 to 39 populations across three conifer taxa: two Douglas-fir varieties and jack pine. We evaluate sensitivity of these algorithms to the source of input loci (markers selected from genotype-environment associations [GEA] or those selected at random). We validate these methods against two-year and 52-year growth and mortality measured in independent transplant experiments. Overall, we find that both methods often better predict transplant performance than climatic or geographic distances. We also find that GFoffset and RONA models are surprisingly not improved using GEA candidates. Even with promising validation results, variation in model projections to future climates makes it difficult to identify the most maladapted populations using either method. Our work advances understanding of the sensitivity and applicability of these approaches, and we discuss recommendations for their future use.
If you use or are inspired by code in this repository, please cite the manuscript:
Lind, B. M., R. Candido-Ribeiro, P. Singh, M. Lu, D. O. Vidakovic, T. R. Booker, M. Whitlock, N. Isabel, S. Yeaman,
and S. N. Aitken. 2024. How useful is genomic data for predicting maladaptation to future climate? Accepted to *Global Change Biology*. Available on bioRxiv DOI: https://doi.org/10.1101/2023.02.10.528022
or code
Lind BM. 2024. GitHub.com/brandonlind/offset_validation: Publication release (Version 1.1.0). Zendodo (2023): DOI: https://doi.org/10.5281/zenodo.10708661
Lind BM. 2023. GitHub.com/brandonlind/douglas_fir_natural_populations: Offset Revision 1 (v1.0.0). Zenodo. https://doi.org/10.5281/zenodo.8018894
Lind BM. 2023. GitHub.com/brandonlind/jack_pine_natural_populations: Offset Revision 1 (v1.0.0). Zenodo. https://doi.org/10.5281/zenodo.8018892
Data archived in conjunction with this manuscript can be found at Dryad using the citation below. Raw sequence reads were deposited on NCBI's Sequence Read Archive under bioprojects PRJNA1079709 and PRJNA744263.
Lind, Brandon et al. (2024). How useful is genomic data for predicting maladaptation to future climate? [Dataset]. Dryad. https://doi.org/10.5061/dryad.sxksn039h
Notebooks are hyperlinked to be viewed on nbviewer.org, but are best viewed in a jupyter session so that cells can be collapsed or scrolled (html and nbviewer versions, links below do not allow scrolling).
Within each notebook is a html drop-down menu that shows python and package versions used. Look for "Click to view module versions" beneath the first cell.
01_split_training_and_testing.ipynb - split populations into training and test sets to use for cross validation of offset methods
02_update_species_range_shapefiles.ipynb - clip future climate (or common garden climate) using species range shapefiles to ease computational burdens of gradient forests
03_clip_climate_data.ipynb - Get range-wide climate and elevation data to fit trained gradient forest models to current/future climate
04_create_datasets-wza_baypass_random.ipynb - create the files that will be needed to run the R script created in ../05_gradient_forest_processing_scripts.ipynb
05_gradient_forest_processing_scripts.ipynb - create R scripts for training and fitting gradient forest models
06_get_common_garden_climate_data.ipynb - retrieve common garden data for the time of growth for douglas-fir and jack pine
07_fit_gradient_forest_models_to_climate_data.ipynb - fit output models to climate of common garden as well as future climate projections
08_climate_and_geographic_distance_vs_phenotypes.ipynb - calculate 68AB climate and geographic distance
09_RONA.ipynb - Calculate genetic offset using risk of non-adaptedness (RONA, sensu Rellstab et al. 2016)
11_validate_GF_offset_predictions.ipynb - validate offset calculations by plotting relationships between GF offset and phenotypes, and calculation correlation coeficients for pearson and spearman
13_consistency_of_factor_importance.ipynb - investigate the consistency of the rank of environmental importance output by Gradient Forest
14_GF_figs_with_RONA_and_climate-geo.ipynb - create figures to visualize validation of Gradient Forests, combine RONA and non-genetic validation scores
15_recreate_table_1.ipynb - create table of locus counts for manuscript using printouts from 04_create_datasets-wza_baypass_random.ipynb
16_future_RONA.ipynb - estimate RONA for future envs
17_future_RONA_and_GF_figs.ipynb - create figures from future RONA, and correlations between future RONA and future GradientForests
18_gf_future_offset_maps.ipynb - create figs that show predicted offset to future climates - choose which model to use based on model validation performance
19_climate_PCAs.ipynb - create PCAs for climate that display the populations as well as the common gardens
20_look_at_validation_relationships.ipynb - look to see if the flip flop between interior varieties from the future offset is also apparent in the projection to the common garden for either GF or RONA
21_relationship_between_kfold_and_full_GF_models.ipynb - explore GF model sensitivity by looking at the relationship between the union of out-of-bag offset predictions with that of the model using all populations
22_recreate_scatterplot.ipynb - recreate the scatter plot between GF WZA full agains increment growth for the cross-variety model
99_common_garden_maps.ipynb - create maps of populations that include the common gardens used
extract_importance.R - From a saved RDS object output from gradient forest training, extract predictor importance and save
gradient_fitting_script.R - Given a trained gradient forest, fit model to input climate data, range_data; save offset netCDF file.
gradient_training.R - Given a set of populations, allele freqs, and environmental data, train gradient forests.
pythonimports in notebooks can be found here: https://github.com/brandonlind/pythonimports