Expanding Prediction To Multiple Viruses

We will attempt to replicate dominant strain prediction to two more viruses:

• HIV1-A
  • This will attempt to predict dominant strain within individual patients
• CCHFV
  • This is an attempt to predict dominant strain over time, say for 5 years

The key ponts here are: 1) is there enough data available to carry out this modeling? And 2) should we consider geographic stratification?

Lets start with CCHF. We do the following steps:

• Download all CCHFV sequences between collection years 2010 and 2020
• breakdown by 2 year periods
• check which segments are most common (there is S, M, and L segements in the virus)
• Try making Qnets with this 2 year data (after aliognment, and segment choice)
• Carry out dominant strain pred and validation (check how well we do from the actual dominant strain prediction)

Currently looking at HIV patient specific data availability

Task List

1. generate the alphabeta and the metadata csvs as above
2. Shere the link with the diffusion simulation u did for the different viruses
3. Dominat strain tracking task (above)
4. We need to do a EPITOPE MATCH task (detailed later)
5. We need to do a IRAT calculation task as well (this will be a seprate paper)

Epitope Match Task

Let k be the diffusion simulation step.

1. compute the MHCI top epitopes using the application u tried, call this set E_k

2. compute the deviation of E_k from E_0 as follows:

   Note these are sets of strings Compute $$average_{x \in E_k} min_{y \in E_0} L(x,y)$$

3. Plot the averge E_k line, averaged over mutiple simulation runs

IRAT task

1. Show qIRAT for 100 H1N1, 100 H3N2, and sufficient number of recent H5 and H7 sequences (report minimum score over 2010-current, and the current year score)
2. List top 3 for each variety (minimum current year score)
3. as usual report seqName, seqAccession, etc