• Search the literature for further mutations to add to the DB

@Sam-Fin might have got these wrong.

add a column and assign a number:
1 - The mutation occurs in 2 studies / 2 viruses (non sars2). Or once in a sars2 study.
2 - seen once but good site homology to sars2
3 - seen once poor homology to sars2.

• Identify what needs changing in the textdb to make it compatible with sneff stuff
• action this: newcol [sarscov2ProtName, sarscov2polyprotaaloc]
• The database is currently in text format, explorer reafactoring to relational. say a table for
  references m21, protein m2m, mutations 12m

For each mutations, add a column with insights.
i.e. this mutations sticks out into nothing, in sars2 -> unlikely to impact drug binding.

• Judy's data
• Add noro stuff
  Text db format for now

A potentially fun and potentially interesting approach to trying to find relevant research papers for this project.

1. Define a epmc search for articles & download as xml
2. Pre-process into a table of id / doi / title / abstract
3. Abstract text -> NLP to remove fluff words -> words2vector
4. Abstract distancees / DimReduction
5. Clustering of abstract vectors
6. Visualisation of search space
7. highlight already identified key papers, see if they cluster if not remove things like the virus words which micght contribute a lot to "virus specific clusters" which we don't want if it splits up our key papers.

OC and AG discuss the solution and implement.

• Identify proteins in the DB
• Generate swissmodel predictions as a first pass
• Generate Alphafold PDB's for these proteins
• Generate a large psi blast of homologous rdrp proteins, so we can look at conservation and pssm's etc

BLOCKED: GPU computer offline for maintenance
now available

@Sam-Fin will generate swiss models structures in the meantime