Treating Influenza: A “Good Old Days” Paper on a Virus That Isn’t Covid

The Paper: Optimization of affinity, specificity and function of designed influenza inhibitors using deep sequencing

The Jist of It: Authors Whitehead et al. (2012) use “deep sequencing” to search for antiviral agents, targeting influenza viruses. In short, they took a protein sequence, mutated it randomly throughout, and looked for new versions that were better at binding to viral proteins. (If an inhibitor protein binds a viral protein, in this case, it deactivates it.)

Approach: Generate a library of random mutations for some protein that binds the flu. Use fluorescent tags to mark how strongly that binding interaction occurs, and use Fluorescence Activated Cell Sorting (FACS) to separate strong binders from weak binders. We can then look at which mutations are present in the strong binders, which are present in the weak binders, and see which mutations are most useful.

Interpreting the Data: We can look at the enrichment of individual protein mutations and see if they show up more frequently in tight binders than others. “Enrichment” is the difference between mutation counts in the selected population and mutation counts in a “reference,” or unselected, population. The unmutated, original protein isn’t a great binder so shouldn’t change much between these two populations. That makes it a good control to compare mutated versions too.

Cool Strategies: If a protein is better at binding some viral proteins than others, it is probably a specific inhibitor. Sometimes we want broadly neutralizing proteins; sometimes we want specific ones. This paper presents a straightforward method for whether a variant does one or the other (or neither).

What Makes the Sequencing Deep? To obtain this data, the researchers had to extra DNA from yeast cells harboring the proteins mutated in the study. This DNA was extracted, and the protein cut out with PCR methods. This was sent for Illumina sequencing, which provided over a million reads for some population. In their study, the typical variant had at least a hundred reads. This allows for a rich dataset informing which variants are enriched when they inhibit a viral protein, and when they do not.