Did you know? You don't lay eggs thanks to an ancient viral infection. The Syncytin gene, crucial for the placenta, was captured from a retrovirus.
APSPM 2026 Workshops
Protein structures are known to be 3x to 10x more conserved compared to protein sequences - allowing for inference of deep homology, which often is not accessible with conventional sequence analysis.
Three beginner/intermediate-level workshops will run in succession on the second day of the meeting (Tuesday 17 Feb; 1PM - 5PM) to train the audience on extracting this conserved signal from protein structures. The workshops will use cutting-edge methods allowing attendees to gain relevant skills and productively contribute to the field.
(Note: You will need a modern laptop, with a modern OS and browser to access the exercise material provided. A detailed list of softwares requirementsd will be communicated closer to the workshop date)
Workshop 1: Structural Homology Detection
Instructors:
TBA
Duration:
75 minutes
Description:
Because of tools like AlphaFold, there are now hundreds of millions of high quality protein structural predictions available online. This workshop will show how Foldseek can be used to rapidly search these databases and find homologous structures.
Workshop 2: Phylogenetic Inference from Structure
Instructors:
TBA
Duration:
75 minutes
Description:
Protein structures are well-conserved over long evolutionary timescales, making them particularly useful for phylogenetics when sequence divergence is high. Using IQ-tree, this workshop will demonstrate how phylogenetic trees can be estimated under the 3Di structural alphabet employed by Foldseek.
Workshop 3: Protein Language Models and Phylogenetics
Instructors:
Ashar Malik
Duration:
75 minutes
Description:
Protein Language Models (pLMs) allow latent space representation of proteins, both sequence and structure. This workshop will explore how this latent space can be used to extract evolutionary signals and put it towards phylogenetic inference. The workshop will introduce pLMs, look at state of the art models like ESM-Cambrian and SA-Prot, show how to extract latent space representations and their respective computational transformation for phylogenetic inference.