Week 2

05/10/2023

Work done

• Literature review MSA sections
• Plan created for possible project direction
• Deep Q-network research and implementation (connect4) for research

Topics for meeting

• How is deep learning RL loss calculated and back-propagated
• What exactly would the state and action spaces be (biggest problem so far)

Meeting Summary

• State space: all possible sequences
• Action space: insert/delete gap at position x,y (insert n gaps?)
• Discussion on Bellman equations
• Possibility of using CNNs to spot motifs
• Discussion on databases to use: UnitProct, BLOSUM, BLAST, Pfam

Things to do

• Create RL agent to add/delete gaps for a small number of small sequences
• Continue literature review