David Enard
@denard.bsky.social
Evolutionary Biologist. Ancient epidemics. Genomic adaptation.
Funny since Judo is in fact a LOT about sweeps and disequilibrium!
January 22, 2026 at 2:04 PM
Funny since Judo is in fact a LOT about sweeps and disequilibrium!
5/5) To learn about normalizing sweeps stats by recombination bins, below is a great paper:
pmc.ncbi.nlm.nih.gov/articles/PMC...
pmc.ncbi.nlm.nih.gov/articles/PMC...
Patterns of shared signatures of recent positive selection across human populations
Signatures of recent positive selection often overlap across human populations, but the question of how often these overlaps represent a single ancestral event remains unresolved. If a single selective event spread across many populations, the same ...
pmc.ncbi.nlm.nih.gov
January 22, 2026 at 1:01 PM
5/5) To learn about normalizing sweeps stats by recombination bins, below is a great paper:
pmc.ncbi.nlm.nih.gov/articles/PMC...
pmc.ncbi.nlm.nih.gov/articles/PMC...
4/5) We also find in preliminary tests that DANN form the Siepel lab really helps mitigating these issues:
journals.plos.org/plosgenetics...
journals.plos.org/plosgenetics...
Domain-adaptive neural networks improve supervised machine learning based on simulated population genetic data
Author summary Population genetic simulation is a powerful tool in the study of evolution. A number of supervised machine learning methods have been developed that take advantage of inexpensive simula...
journals.plos.org
January 22, 2026 at 1:01 PM
4/5) We also find in preliminary tests that DANN form the Siepel lab really helps mitigating these issues:
journals.plos.org/plosgenetics...
journals.plos.org/plosgenetics...
3/5) real genome. A stat like iHS for example, or anything haplotype-based really. We are working on adding this to Flexsweep and preparing a manuscript, but even before that anyone using CNNs for sweep analysis should be very careful about recombination.
January 22, 2026 at 1:01 PM
3/5) real genome. A stat like iHS for example, or anything haplotype-based really. We are working on adding this to Flexsweep and preparing a manuscript, but even before that anyone using CNNs for sweep analysis should be very careful about recombination.
2/5) , and/or normalizing statistics by local recombination rate is really important. This is because many sweep stats values depend on local recombination rate, and thus normalization based on genome-wide variants will be heavily dependent on the recombination simulated in the training vs. in the
January 22, 2026 at 1:01 PM
2/5) , and/or normalizing statistics by local recombination rate is really important. This is because many sweep stats values depend on local recombination rate, and thus normalization based on genome-wide variants will be heavily dependent on the recombination simulated in the training vs. in the
I use it every year at the start of my popgen course, asking the students to think about how they can explain what they see.
January 10, 2026 at 2:55 PM
I use it every year at the start of my popgen course, asking the students to think about how they can explain what they see.