Skip to content

Week 09: Gradient Forest: SDMs for genomic data

Use SNP genotypes and climate rasters to estimate environmental correlates of genetic turnover in P. guttatus.

Icebreaker

Who is your favorite cartoon character and why?

Paper discussion

Cancelled for this week

Core Questions

  • How can we apply an SDM framework to study changes in genetic variation within species as a function of environmental variation?
  • How does Gradient Forest combine SNP-specific random forests into environmental turnover curves?
  • How can we combine all the information from SNP-specific predictions to understand genome-wide turnover as a function of environmental variation?

Concepts

  • SNPs, VCF genotype calls, and individual alternate-allele dosage (0, 1, 2)
  • One random forest per SNP
  • Cumulative turnover curves and environmental thresholds
  • PCA of transformed environments; shared map/biplot colors and environmental covariance arrows

Python Tools

  • pyGF
  • GeoPandas, pandas, NumPy, xarray/rioxarray, and Matplotlib

Applied Lab

Pull the latest course repository on JupyterHub.

Students read the P. guttatus VCF and sample coordinates, filter and sample up to 256 SNPs, and extract 19 bioclimatic predictors. They fit SNP-specific forests, inspect OOB scores and predictor importance, interpret cumulative turnover curves, and map GF-transformed environments with a shared-color PCA biplot. They compare the aggregate pattern with a single-SNP view and export diagnostics, turnover coordinates, figures, and a GeoTIFF with one band per predictor.

Assignment

Repeat the Lab09 workflow for another snake species and submit the completed assignment notebook, diagnostics, turnover curves, maps, and exported turnover raster. Identify an important predictor and the part of its environmental gradient with the fastest modeled turnover. Compare the aggregate and single-SNP biplots, explaining why their colors cannot be compared directly. Discuss one limitation of OOB evaluation and one alternative to local adaptation that could explain a genetic–environmental association.

Paper discussion leader next week: Braedon! Please select a paper for the group to discuss before Sunday 11/01 6pm and send it around to the class email list: bio597-fall2026-group@maine.edu