RADCamp Latin America 2026 Part II (Bioinformatics)

Day 3 (AM)

Now that your assemblies have completed, what are the next steps?

Inspect the assembly
Clean it up a bit (if needed)
Downstream analysis

Overview of the morning activities:

Examine quality of the assemblies
ipyrad2 Export tools
ipyrad2 analysis tools
ipyrad analysis API - Principal Component Analysis (PCA)
Coffee Break (10:40-11:00)
Phylogenetic inference w/ treeslider & Astral

Examine quality of assemblies

Lead: Isaac (30’)

Lets take some time to look at results of the assembly process for some of the real datasets. Did they all work perfectly? Why did some work, why did some break? Look at runtimes: How long did they take to run?

Look at the Step 5 stats file and understand what it’s telling you.
Use ipyrad inspect to view assembly results: ipyrad2 inspect <assembly_name>_outfiles/

Here are some questions to scaffold your exploration of your assembly. Write down the answers to these questions and be prepared to discuss them with the group:

Do all the samples have a similar number of loci in the final assembly?
Are there any samples you would choose to remove? If so which are they and why?
TODO: Add more questions for evaluating the empirical assembly quality

ipyrad2 export tools

Lead: Deren (30-45’) In this exercise we will export a phylip file and run/plot a quick phylogenetic tree as a ‘reality check’ of the assembly results.

Filtering and writing outputs w/ ipyrad analysis export tools
Talk about mindepth/minsamp, removing samples
Evaluate and remediate missing data (if needed)
Run a quick tree in raxml
Plot the tree with toytree

ipyrad analysis tools

A brief overview of ipyrad analysis tools.

$ ipyrad2 -h

analysis subcommands
    pca                                      Infer population structure from pca, tsne, or umap on filtered SNPs
    dapc                                     Infer population genetic clustering by discriminant analysis of principal components
    snmf                                     Infer population genetic clustering by non-negative matrix factorization
    admixture                                Infer population genetic clustering with external ADMIXTURE
    popgen                                   Infer population genetic statistics for one or more populations
    bpp                                      Infer species tree; species delim; or MSC+ model from multi-locus data
    baba                                     Infer admixture metrics from ABBA/BABA and related SNP patterns
    treeslider                               Infer gene trees for each qualified locus or refmapped genomic window of loci

ipyrad analysis API - Principal Component Analysis (PCA)

Lead: Isaac (30’)

Exercise: ipyrad analysis API for PCA

Coffee Break (10:40-11)

Phylogenetic inference with treeslider and astral

Lead: Deren (TODO: duration)

Brief intro to other analysis tools

ipyrad2 analysis tools documentation

sNMF

A tool for quantifying population structure similar in spirit to other populare tools like structure or admixture but much faster. You can use ipyrad2 snmf to get the estimated best K number of populations and also ancestry coefficients for each sample under this K value (which you can use to make structure plots).

Here is an example run for the simulated data (ipyrad2 snmf -h shows all options):

ipyrad2 snmf -d peddrad_outfiles/peddrad.hdf5 -o peddrad_snmf --k-range 2:5 -c 8

And the relevant output files:

snmf.k_scan.tsv

This is the file that shows the best K value (in selected column) and the average cross entropy scores across the given number of replicates that were run (lower mean_cross_entropy is better).

k       mean_cross_entropy      sd_cross_entropy        best_reconstruction_err best_n_iter     selected
     2.1157303302813344      0.08450922931622941     43.15841209164097       41              False
     2.038834028432327       0.20176281049999648     37.86938961917523       57              True
     2.4625634376913736      0.23786036992865084     34.69336020004499       53              False
     2.5947746558961002      0.22351698368333492     31.423331464304972      71              False

snmf.membership.tsv

This is the ancestry proportion assignment for each sample under the best K number of populations (similar to the structure Q matrix). You can use this file to make structure plots showing ancestry fraction per sample.

sample  cluster1        cluster2        cluster3
1A_0    1.0             0.0             0.0
1B_0    1.0             0.0             0.0
1C_0    0.86978335585   0.075848657     0.05436798704612854
1D_0    0.72944634665   0.192959939     0.07759371369923207
2E_0    0.0             1.0             0.0
2F_0    0.0             0.989428831     0.010571168524944852
2G_0    0.0050992874    0.96519540      0.029705304501056785
2H_0    0.075780730     0.8589390       0.06528018397030126
3I_0    0.0             0.0             1.0
3J_0    0.0             0.0             1.0
3K_0    0.0             0.0             1.0
3L_0    0.021929335     0.035726021     0.9423446429693687

popgen

A tool for running a suite of several standard population genetics summary statistics including within and between population statistics. By default you will get pi, dxy, fst, tajima_d, theta_w, heterozygosity, fis, fit, sfs.

Here is an example run for the simulated data (ipyrad2 popgen -h shows all options):

ipyrad2 popgen -d peddrad_outfiles/peddrad.hdf5 -o peddrad_popgen -i sim_pops.txt -c 8