Page 1 of 3

Exporting substrate statistics

Posted: Thu Feb 02, 2017 10:33 pm
by Petter
I've been thinking of adding a checkbox to the advanced substrate settings dialog that says something like "Write simulation stats to file cell_lab_stats.csv". When checked, a line would be written to that file every one simulation hour or so. That line would start with the current simulation time, then perhaps the total number of cells, or maybe the number of cells in each mode, or perhaps of each cell type?

Any thoughts on this? What are relevant stats to write? Don't want to write too much of course and don't want to make it customizable, just find like a reasonable set of good statistics to write.

This file is then meant to be imported to e.g. excel where one can study e.g. population oscillations, adaptions and long term trends. Just fun to plot stuff and might be a bit educational. It's though meant to be kind of an advanced use so users will have to figure out how to use it best themselves. Maybe someone can make an app to plot it one day :P.


Re: Exporting substrate statistics

Posted: Thu Feb 02, 2017 10:56 pm
by Nayus
That would be really useful! :) That would give good data for comparisons and studies

About the relevant stats: Cell Count, mainly. Food amount. I think we'll all agree on those ones.

I don't know if this'd be really useful, bit I'm thinking that would be good to see what exactly is killing or helping the organism. This could be shown by something like "death by starvation" and "death by age" counters. I don't know if there's a way to count this difference, or if there could be actually useful. Just to let it out here.

Other way of counting this that would give other uses as well could be a individual counter of each type of cell. Like a counter of how many Devorocytes are in the substrate in any given moment. So if in the subtrate an organism has only one Devorocyte, for example, you can count how many organisms are there without a complicated algorithm.

But that maybe would be too much data to count per hour. I don't know.

Re: Exporting substrate statistics

Posted: Thu Feb 02, 2017 11:04 pm
by Petter
Cell count and food amount are a given. Though we should split the cell count so that we can distinguish different species. The question is how. Every mode could work, but what if some species use all modes/same modes? Per type might be better but might still not be enough. Per color could work, like 4^3=64 columns. In principle a column for every mode and type combo is possible but that would be 200 columns and that might be a bit much.

What you wrote about deaths is interesting, a bit too much work for me given the usefulness i think since most often when optimizing cells die of lack of food i guess. However the number of splits and deaths per hour at that moment would be interesting.

Re: Exporting substrate statistics

Posted: Thu Feb 02, 2017 11:38 pm
by wapcaplet
I would love to have a CSV stats feature of some kind, especially the idea of showing cell count for each species. I've often wanted to see the total population of two or more genomes on the same substrate, for example to monitor predator/prey or compare several variants of the same organism. By mode or cell type would be less useful, I think; it would be a ton of data that I'm not sure would tell us much.

By color might be interesting, though it seems infeasible to log data for every color. Maybe if you could hard-code cell count stats for a handful of distinctive colors (say, pure red/green/blue/yellow/cyan/magenta), we could color-code our own genomes to get the stats we're looking for without a huge number of columns. For example I could make one genome all red, and another all blue, then compare the cell counts for red and blue. Or I might make my devorocytes cyan, and phagocytes magenta, and compare their populations based on the color stats.

Re: Exporting substrate statistics

Posted: Fri Feb 03, 2017 5:40 pm
by bwisialo
There's no straightforward way to do cells per genome? I would have thought this was how counts of "Initial cells" vs "Your cells" worked.

Re: Exporting substrate statistics

Posted: Fri Feb 03, 2017 6:02 pm
by Petter
On a challenge substrate there is a clear distinction between those that were there from the start and those that the user added. But how would this work on an experimental plate? I want a general solution that in principle could track the populations of three simultaneous species. Different colors could e.g. work then. Tracking descendants of each inserted cell could work too, with numbers appearing in order of insertion. Any concrete suggestions are welcome. Turning on contaminate would immediately increase this number to the thousands though...

Re: Exporting substrate statistics

Posted: Fri Feb 03, 2017 7:08 pm
by bwisialo
Got it. Yeah, I realized that cell per genome would get ridiculous when applied to plates in the midst of evolution. :lol:

Re: Exporting substrate statistics

Posted: Fri Feb 03, 2017 7:26 pm
by bwisialo
Okay, I'd argue that Cell by color stats are best here.

If one genome is all one color, and another genome is all another color, then cell per color gives you cells per genome. If each cell/mode in a genome is a different color, and genomes don't have color overlap, then you can add the sums to get cells per genome; the individual colors will in effect give you cells per mode for each genome, which can also be used to get cells per type in each genome.

Color coding your modes is very easy. If you started with Cell by mode stats, mode overlaps will prevent you from being able to figure out the breakdown described above. Mode overlaps are much more likely and much less easy to keep under control than color overlaps.

Getting clear info from a mid-evolution plate will be data-overload and extremely confusing no matter what.

Re: Exporting substrate statistics

Posted: Fri Feb 03, 2017 7:34 pm
by Petter
Ok, I'm thinking splitting each color into 4 bands, and then having 4*4*4=64 different color categories that we track the number in.

What about giving each human inserted cell an id starting at 0 (regardless of identical genomes). All cells inserted from contaminate are given id -1. Mutations do not change id.

I feel like we might need a vote in a few days :P

Re: Exporting substrate statistics

Posted: Fri Feb 03, 2017 7:38 pm
by Petter
Another option could be tracking color of a certain mode only, like M1. Then all cells in an organism would always have the same category which i guess is desirable. Just 3 bands might be enough though, 3*3*3=27 categories. Only issue is that mutations might change a cells category. But with 27 categories it's still kind of unlikely to have a clash.