I had the "privilege" of reviewing a certain piece of software written to model a certain pandemic outbreak...
Numerical modelling in biology/virology feels rife with serious problems that are simply sidestepped because of ego and status. I raised alarms and as merely a senior expert in distributed computation and numerical modelling I was shutdown quite strongly. After the group in question were forced to accept our published work refuting their results to be FUBAR due to severe numerical instabilities which hadn't even been checked for we were shut out for a group who patted them on the back for being "oh so clever"...
(our published work numerically demonstrated results they were presenting as fact were compatible with statistical noise and that it is known they didn't have the required extra 100xcompute or 100xtime to produce results as precise as what they were after)
Had a similar experience a few years ago with a geological modelling group and a group hoping to do biochemistry on a huge cluster of 20k cpu cores.
For all of the "numerical sciences" achievements only really physics and chemistry (and maths) seem to have championed the ideas of reproducibility and falsifiability in numerical analysis and simulation.
Numerical modelling in biology/virology feels rife with serious problems that are simply sidestepped because of ego and status. I raised alarms and as merely a senior expert in distributed computation and numerical modelling I was shutdown quite strongly. After the group in question were forced to accept our published work refuting their results to be FUBAR due to severe numerical instabilities which hadn't even been checked for we were shut out for a group who patted them on the back for being "oh so clever"...
(our published work numerically demonstrated results they were presenting as fact were compatible with statistical noise and that it is known they didn't have the required extra 100xcompute or 100xtime to produce results as precise as what they were after)
Had a similar experience a few years ago with a geological modelling group and a group hoping to do biochemistry on a huge cluster of 20k cpu cores. For all of the "numerical sciences" achievements only really physics and chemistry (and maths) seem to have championed the ideas of reproducibility and falsifiability in numerical analysis and simulation.