use case: John Snow

[phiweger]

• get map Soho and approximate location of the cholera cases he registered
• generate a tree
• the data for that comes from sequence data unrelated to this outbreak, they are just there for illustrative purposes
• map the cases and query by location (we suspect a pump, how many cases in the perimeter of it)
• each pump has a count of cases: construct a Bayesian model (cholera_count ~ poisson(...)) that identifies the right pump (that is the modelling/ machine learning part)
• the data of the outbreak is "discovered" by streaming raw read sets (metagenomic water/ stool samples) and checking for cholera, they exist only as unannotated data dumps w/ minimal annotation and geotag (we anticipate here the rise of the minisequencer + app in smart phone, assumption: people will be reluctant to place data in any special format (such as ENA ...)

[phiweger]

or sift through human stool datasets for Vibrio cholerae: if match, assemble, analyse