Master Practicum: Applying Deep Learning on Proteins in Post-AlphaFold2 Times

Course Info

Type: Master Lab

ECTS: 12

Supervisors Prof. Burkhard Rost, tba

Rotation: (prospective; TBD with students) weekly Zoom meetings

Language: English

Topics, where and when

Kick-off meeting will be at the beginning of the summer term (we will update this page! Check back here!)

All of our projects involve deep learning techniques and experience with PyTorch may be an advantage.

Possible topics (preliminary):

  1. Prediction of AlphaFold2's reliability score, pLDDT, based on sequence vector representations (embeddings)
  2. Prediction of AlphaFold2's paired reliability score, PAE, based on sequence vector representations (embeddings)
  3. Investigate usefulness of Alphafold2's paired reliability score, PAE, for domain boundary prediction
  4. Re-create the HSSP curve in the Post-AlphaFold2 time using AlphaFold2 predictions
  5. Predicting binding residues from sequence vector representations (embeddings) and structure representations (distance maps)

The detailed definition of topics and the schedule will be finalized in coordination with the participants.