Example with sample data

This is a step by step tutorial to fit the model A+ST+Gt+CTt = ln(RFP/CFP) to a specific dataset. If you whish to read the general case you should go to the home of the site

You can follow this example to fit measurements from 2 plates of aging fluorescent co-cultures to the model Ax+ST+Gt+CTt = ln(RFP/CFP). First you should download sample data BgDataAllTest.

All of the functions are in Github_AbrahamAvelar. Most of them are in this folder and some others in this other folder

Next thing is to load the sample dataset it into your Matlab workspace

load BgDataAllTest

See what’s in this structure

>>BgDataAllTest

BgDataAllTest = 
1x2 struct array with fields:
    OD
    CFP
    RFP
    t

The most important variable loaded at this point is the structure BgDataAllTest. While the variable ‘muts’ has the positions for each mutants in this exaple, you must make sure that the BgDataAllTest structure has the relevant fields OD, CFP, RFP, and t. Your structure may have other fields but it must contain these four. CFP is the color of the reference strain in WT+mutant competitions.
If you have a structure with other names in the fields, the function bgdataAll2BgDataAll may be useful.

Before fitting the dataset to the model, we get the exponential phase measurements.

refs = muts.refs %We have the positions of reference competitions stored in this variable
plt = 1:2; %because this example dataset has plates 1 and 2.
Tiempo0 =1; %It must be 1 for lineal modeling fitting of relative survivorship analysis
showfig=1;
ExpBgDataAll =  ExtractExponentialPoints(BgDataAllTest, plt, showfig, refs, Tiempo0 )

If you used showfig = 1 then one figure per plate should be printed to the screen. lines show all of the measurements and empty circles in different colour show the selected measurements that were extracted by the function.

Then we compute two time-related variables:
Tdays - how many days passed from the first measurement to the start of each outgrowth, this will be the coefficients T in the equations to fit the model
tOut - It is a time measurement of how long passed from the moment of the inocule of the outgrowth to each read. Is t in the model. Although it would be more intuitive expressed in hours, it is in days.

odTh=0.22;
ExpBgDataAll = calculaTiempos(ExpBgDataAll, plt, odTh)

Then we subtract the background signal from wells that only have the WTrfp or WTcfp strain.

%% Prepare input variables
exp=1; % We put 1 because we have a structure only with exponential points which is the output of 'ExtractExponentialPoints'
bgdataPrueba=BgDataAll2bgdataEGG(ExpBgDataAll,plt,'CFP','RFP',exp); 
OnlyMutStrain = [89]; %Index or Indexes of the reference wells that have only WTcfp
OnlyRefStrain = [96]; %Index or Indexes of the reference wells that have only WTrfp (the same FP as all of the mutants)  
% Execute background subtraction script
BgDataSinFondo = restarFondoFPs(bgdataPrueba, plt, OnlyMutStrain,OnlyRefStrain)

Then we are ready to call the fitting function regress which is inside the ModelASGC script.
To do so this script first builds the predictor matrix and response vector taking into account which wells are WTrfp+WTcfp reference competitions, which have only one fluorescent strain and that they have enough valid measurements. After fitting the data to the model it organizes variables A, S, G and C and some statistics to the output structures.

% Prepare input variables
medicionesminimas=[15]; %Smallest number of valid measurements per well to be included in the fitting function
extraPlRefs = 0; % The number of an extra plate that has references in it. Put 0 if you have references in every plate.
% Execute Linear Model fitting script
[bgdataAll_LM,data2_LM]=ModelASGC(BgDataSinFondo,plt,refs,OnlyMutStrain,OnlyRefStrain,medicionesminimas,exp,extraPlRefs)

bgdataAll_LM contains the coefficients A, S, G and C for each well, then you can find out whether your mutant strains increased or decreased lifespan.

In this example we show what happened to nine mutants whit 6 or 7 biological replicates each.

%% See S to make sure it makes sense
campos = fieldnames(muts);
clf
for pl = plt
    subplot(2,1,pl)
    plot( [0 length(campos)+1], [0 0], '-', 'color', [.5 .5 .5]) 
    hold on
    for i=1:length(campos)
        thisY = bgdataAll_LM(pl).S(muts.(str2mat(campos(i)))) ;
        JitterPlot( i, thisY,  .2, '.', 'k')
    end
    xlim([0 i+.5])
    ylim([-.3 .2])
    set(gca, 'xtick', [1:12], 'xticklabels', campos)
    ylabel('S, relative survivorship')
    title( strcat('Replicate', num2str(pl)) )
end

You should have a plot with all the measurements of S with each one of the plates in one subplot. If you do, you are ready to try it with your own data!.
The script sets al reference wells to NaN but by deffinition of the model they are Sref=0 and Gref=0.

LinearModelCLS

Example with sample data