PRIMAGE Requirements

[Robadob]

Just running through the PRIMAGE code, identifying features not currently supported.

• Extra Cellular Matrix Grid (Cellular Automata) (https://github.com/FLAMEGPU/FLAMEGPU2/issues/308, https://github.com/FLAMEGPU/FLAMEGPU2/issues/162) This grid represents a discretised form of the environment the cells exist in. It holds data related to cell density, o2, matrix value, As it represents the environment, it is not well suited to message communication, as cells frequently read their current grid cell's data (would require multiple message read). Cell's are required to atomically increment population counters at various points, these are then subsequently read (and reset to 0) to calculate dependent cell values (e.g. oxygen consumption). Cells can increment multiple counters, dependent on their internal state. (Some of) These values are required to be recalculated after each iteration of the migration submodel, and after the force resolution submodel completes. This could be achieved with a bunch of separate agent functions for counting cells and the dependent values are various points. But to minimise impact, instead the atomic operations have been merged into the end of preceding agent functions. This isn't too difficult but becomes complicated quickly. In the latest update of the model I have added more thorough commentary to each grid array to try and clarify how it is used.

But really we mostly require a large environment array which represents the environment (and can be acted on). In future, diffusion may also need to be applied to this array. Technically, this grid also grows with the size of the tumour. However, that merely means some cells remain uninitialised until later into the simulation (whereby they copy their neighbour's data).

• Bitfield Variables Not strictly necessary, but these would help reduce the memory footprint of agents as the model contains many bool /3-state agent variables.

• Mode specific outputs Primage model has several different modes, specify a particular runtime argument and file name, and it will output specific output data to that file/directory. This is used for; validation timeseries, visualisation timeseries, force resolution tracking, official primage execution and some other less frequent outputs. Perhaps a way to have runtime toggled exit/step functions (and a way to configure custom json logging).

• Trivial counters/logging within agent functions (#178) I've used atomics quite heavily when validating some methods with complex branching structures, to ensure the correct branches are being followed in the correct proportions. Some equivalent way of doing this would be very useful, especially if it were accessible to python models. This also links in with the mode specific outputs thingy above.

[Robadob]

Made an overview of how Primage env Grid is used here: Note, it doesn't (currently) collide with messaging.

https://gist.github.com/Robadob/fa846e6430eaf645a475f8c3043aa0fb

[Robadob]

Re: Init layers

• Get kenneth to restructure model so that init is absorbed by the regular model (and steps in a different position)

Re: Grid

• Cells output message
• Grid Agent summs messages and outputs grid message with data
• Agents read in grid message when required.
• (This should currently be supported)

[Robadob]

Primage is finished.