Steady-state sensitivities in Matlab: sllh issue

[aidinbii]

Hi, could you please help me out In Matlab when I run steady-state sensitivities in the case of adjoint sensitivities

options_ss.sensi_meth = 'adjoint';

sllh vector has all zeroes.

while in forward sensitivities

options_ss.sensi_meth = 'forward';

it calculates non-zero values

I provide random data

D.Y = [rand(1,30)*50];
D.Sigma_Y = rand(1, 30);

[FFroehlich]

Thanks for bringing this up, this is probably some issue that options are not passed correctly in the matlab interface.@plakrisenko could you please have a look at https://github.com/AMICI-dev/AMICI/blob/master/src/interface_matlab.cpp and see whether we missed adding some of the options to the matlab interface that were added to the python interface?

[FFroehlich]

Hi aidinbii, are you sure that the integration was actually successful, i.e. that no integration error occured?

[aidinbii]

@FFroehlich, yes, the posteq_status is -4; 1; 0 (for both cases: forward & adjoint)

[FFroehlich]

okay, could you please try with the latest version of amici (0.17.0) that was just released and, if the error persists, upload an example script that reproduces the error?

[aidinbii]

@FFroehlich , it did not help

Here is the script:

clc
clear
close all

%% get tuned parameters
load parameter_Set_healthy_ini_cond_SS_SCRNA_new

p = exp(parameters.MS.par(:,1)); % exp bc they are in log scale
k = initial_condition;

%% Simulations

% indices of inactive cells
inact_ind = [1;7;16;20];

% indices of active cells
act_ind = [2;3;8;9;11;14;17;21;22;23];

% indices of proteins
prot_ind = [4;5;6;10;12;13;15;18;19;24];

%%
t = linspace(0,1000,3000);
options_simul = amioption('sensi',1,'sensi_meth','adjoint', ...
    'maxsteps',1e6);
modelName = 'simulate_IgG4_ver1_SCRNAseq_MM1';

%% Random data
D.Y = [rand(1,10)*50];
D.Sigma_Y = rand(1, 10);

%% test adjoint sensi.
out = searchSteadyState(modelName, p, k, options_simul, D);

where the function searchSteadyState:

function sol = searchSteadyState(modelName, param, k_ss, options_ss, varargin)
t = Inf;

[~] = which(modelName); % fix for inaccessability problems

model = str2func(modelName);

if nargin > 4
    D_ss = varargin{1};
    sol = model(t,param,k_ss,D_ss,options_ss);
else
    sol = model(t,param,k_ss,[],options_ss);
end

%% Diagnosis
disp(['The status is ', num2str(sol.diagnosis.posteq_status')]);
disp(['Timepoint reached ', num2str(sol.diagnosis.posteq_t')]);
end