Upload all MorphEUS scripts

.gitignore

+legacy/R/.Rhistory
+*.csv
+*.m~
+*.xlsx
+*.asv
+*.fig
+classification_and_analysis/trial_results/*
+classification_and_analysis/logs/*
+version_info.md
+
+\.DS_Store
+.DS_Store
+
+*.eps

README.md

-# morpheus
+# MorphEUS 
 
-Code for Morphological profiling of tubercle bacilli identifies drug pathways of action
\ No newline at end of file
+Contains two main directories:
+
+## segmentation_pipeline
+
+Inside this directory is the code used to transform the .csv files outputted by MicrobeJ into a MATLAB workspace. This code filters out of focus and blurry cells using the transverse profile and calculates the median, Q1, Q3 and IQR for each replicate.
+
+
+## classification_and_analysis
+
+Inside this directory is the code used to run the classification trials as outlined in figure S3 as well as code to create the consensus KNN (cKNN) and heatmaps to visualize the results from the classification trials. 
+
+In addition, code is included to perform the Kruskal–Wallis tests to determine signfiicance of variables based on using single-cell data and not replicate averages as seen in Figure 1. There is also code to create PCAs that were used in the earlier versions of the analysis pipeline (as seen in Figures 2 and S1). 

classification_and_analysis/PCA_creator.m

+%% PCA Analysis with user prompts for workspace, number of variable, and drug selection
+% DESCRIPTION:
+%     Prompts users to obtain the following variables
+%         - numvar
+%         - chosen_workspaces
+%         - chosen_drugs
+%         - drugs_to_apply
+% 
+%     Having declared these variables, runs the following scripts in this
+%     order:
+%         - load_workspaces
+%         - PCA_analysis
+%
+% PREREQUISITES: none
+% 
+% OUTPUT: ~fill this out~
+
+%% Clear workspace before start
+%~~ Comment out to disable this feature ~~%
+%clear
+
+%% Add Paths
+add_all_paths 
+
+
+%% Prompt for figure workspace
+
+%figure_folder_path = inputdlg("Folder to save figures in:", "Figure Path", [1 35], "./figures/");
+
+figure_folder_path = "./figures/";
+%% Prompt for PCA_analysis settings
+
+% Scripts like get_existing_feature_set set these variables, and we want
+% them to be respected in the gui selection. If they're no set there
+% though, they still need a default value.
+if ~exist('do_feature_reduction','var')
+    do_feature_reduction = true;
+end
+
+if ~exist('remove_extra_controls','var')
+    remove_extra_controls = true;
+end
+
+disp('bad treated = IMI, Dau, Nig, Nal');
+variables_to_create = {'disc', 'do_feature_reduction', 'include_DMSO', 'show_clustergram', 'show_pca', 'plot_3', 'plot_3_DMSO', 'plot_before_tvn', 'remove_INH_control', 'remove_extra_controls', 'remove_bad_treated', 'prompt_for_one_dose', 'use_medians_only'};
+default_values =      [true,   do_feature_reduction,      false,            false,            true,       false,    false,       false,            true,                 remove_extra_controls,     true,               false,                  false];
+
+target_values = checkboxList("Select PCA_analysis Settings", variables_to_create, default_values);
+
+initialize_variables %Initialize variables_to_create as target_values
+
+if use_medians_only
+    numvar = 25;
+    
+end 
+
+%% make this global to make it accessible inside ver_figsave()
+global chosen_workspaces
+
+%% Select number of variables to use
+
+if do_feature_reduction
+    prompt = {'Enter number of variables:'};
+    dlgtitle = 'Variable Number';
+    dims = [1 35];
+    definput = {'94'};
+    numvar = inputdlg(prompt,dlgtitle,dims,definput);
+    numvar = str2num(numvar{1});
+else
+    if exist('pca_whitened_table','var')
+        numvar =  sum(varfun(@isnumeric,pca_whitened_table,'OutputFormat', 'uniform'));
+    else
+        
+    end 
+end 
+%% Find list of workspaces, prompt user for choice, and load chosen
+if ~exist('workspace_directory', 'var')
+    workspace_directory = './workspaces';
+    disp("Defaulted workspace_directory to './workspaces' in load_workspaces")
+end
+
+workspace_list = get_workspaces(workspace_directory);
+if isempty(workspace_list)
+    error(strcat("No workspaces found in ", workspace_directory))
+end
+% If we are operating from an existing feature set, automatically select
+% the workspaces that were used to calculate that set
+if exist('overall_vars', 'var')
+    try
+        default_selection = find(ismember(workspace_list, overall_vars.WORKSPACE));
+    catch
+        disp("Note: overall_vars does not have a WORKSPACE field")
+        default_selection = [];
+    end 
+else
+    default_selection = [];
+end
+
+% prompt the user to select 
+[indexes_chosen, any_chosen_tf] = listdlg('ListString',workspace_list,...
+    'Name',"Select workspace",'ListSize',[380 380],...
+    'InitialValue', default_selection);
+
+% if the user did not select anything, exit (there is a catch for this in load_workspaces too)
+if ~any_chosen_tf
+    return
+end
+
+% use the index to get the list of user-selected workspaces
+chosen_workspaces = workspace_list(indexes_chosen);
+
+%% run the script to load the workspaces
+load_workspaces
+
+%% From the list of drugs, prompt user to select which drugs to use and apply
+
+% If we are operating from an existing feature set, automatically select
+% the variables that were used to calculate that set
+if exist('overall_vars', 'var')
+    default_selection = find(ismember(choice_variable, overall_vars.DRUGS));
+else
+    default_selection = [];
+end
+
+suggestion = {'Mer','Amp','Ctax','INH','EMB','ETA','IMI','Van','Cyc','Del',...
+'Lev','Mox','Clz','MIT','Olf','Kan','Amk','Cam','Cla','Dox','Gent',...
+'Strep','Tet','Lin','Pre','CCCP','Cer','Mon','Nig','Thi','RifT','BDQ',... 
+'RIF','THL','water','Untreated'};
+
+if exist('suggestion','var')
+    default_selection = find(ismember(choice_variable,suggestion));
+    
+end 
+
+% For extra_resolution_confusion
+if prompt_for_one_dose
+    select_one_dose
+end
+
+% Prompt user to select a list of drugs
+[indexes_chosen, any_chosen_tf] = listdlg('ListString',choice_variable,...
+    'Name',"Select Drugs",'ListSize',[220 380],'PromptString',"Select Drugs",...
+    'InitialValue', default_selection); 
+
+% if the user did not select anything, exit
+if ~any_chosen_tf
+    return
+end
+
+% Get a list of the drugs chosen
+chosen_drugs = choice_variable(indexes_chosen);
+
+%prompt the user to select drugs to apply if they want to 
+[indexes_chosen, any_chosen_tf] = listdlg('ListString',choice_variable,'Name',"Select Drugs to apply",'ListSize',[220 380],...
+    'CancelString','No Selection'); 
+
+% Get a list of drugs to apply
+
+drugs_to_apply = choice_variable(indexes_chosen);
+
+%% Run PCA Analysis
+
+PCA_analysis
\ No newline at end of file

classification_and_analysis/README.md

+# MorphEUS Classification Trials and cKNNs
+
+Performs a classification trial on a workspace created with MicrobeJ_segmentation
+
+Currently set to use default workspaces used in the MophEUS paper, can be adjusted.
+
+Run the script classification_trials.m to run a set of 70 classification trials. Results will be automatically saved into trial_results with the name the user selected along with a timestamp. A log containing all of the output produced during hte run will be saved into the logs folder. 
+
+### Before you run
+
+In order to run a classification trial, you need to download Hanchuan Peng's mRMR feature selection algorithm and compile it to work on your operating system. It can be found here:
+[mRMR Feature Selection] (https://www.mathworks.com/matlabcentral/fileexchange/14608-mrmr-feature-selection-using-mutual-information-computation?s_tid=prof_contriblnk). Place the resulting mRMR_0.9_compiled folder in the classification_trials directory.
+
+
+### Workspaces for current presets
+
+Joint profile: full_025x_and_3x_redos_no_A22.mat
+
+Timecourse applying: all_timecourse_doseresponse.mat
+
+High dose: full_3x_with_redos_no_A22_drug_name_only.mat
+
+Low dose: full_025x_with_redos_v3_drug_names_only.mat
+
+## Examining the data
+
+To look through a set of classification trials, run cKNN_creator.m and select the desired workspace. 
+
+To create boxplots using cell data created with MicrobeJ_segmentation, run boxplots_and_KW.m and selecte the desired workspace(s).
+
+To create a PCA, run PCA_creator.m. 

classification_and_analysis/add_all_paths.m

+%% quick script to add all subfolders to path
+
+addpath('./analysis')
+addpath('./application_loops')
+addpath('./functions')
+addpath('./helper')
+addpath('./mRMR_0.9_compiled')
+addpath('./workspaces')
\ No newline at end of file

classification_and_analysis/analysis/KNN_analysis.m

+%% Script to perform KNN on data outputted by PCA_analysis
+% This script uses existing variables in the workspace to perform KNN
+% and output grahps
+%
+% PREREQUISITES:
+% - ** Must have run PCA_analysis.m and all of its prerequisites **
+%       - Most important variable is scores_table
+% - All required settings in section below
+% - Any desired optional variables listed below
+%
+% OUTPUT:
+% - 
+% 
+% DEPENDENCIES:
+% - knn_helper ***
+%
+%% Required setting variables
+%%% Must include all of these variables in the workspace before running %%%
+%
+% - make_median - boolean whether to make a plot of replicate medians
+% - make_full   - boolean whether to make a plot with each replicate as a point
+% d_metric, create_graph, create_cm, large_group
+
+%% Optional settings
+%%% Optionally include these variables in the workspace before running if %
+%%% desired. If not present, they will default to the specified value     %
+%
+% - random_compare - (default: false) - make a randomized label plot to compare
+% - tit_add        - (default: "")    - 
+%% Add paths
+addpath('./functions')
+addpath('./helper/')
+
+%% Generate defaults for optional settings
+if ~exist('random_compare', 'var')
+    random_compare = false;
+    disp('defaulted random_compare to false');
+end
+
+if ~exist('tit_add', 'var')
+    tit_add = "";
+    disp('defaulted tit_add to ""');
+end
+
+if ~exist('confuse_controls','var')
+   confuse_controls = false;  
+end
+
+if confuse_controls
+    make_median = false;
+    make_full = true;
+    
+end 
+
+disp("begining of knn_analysis, value of remove_after_TVN")
+disp(remove_after_TVN)
+
+%% Validate required settings
+
+%From PCA_analysis -- just check for these two; they're the most important
+if ~exist('scores_table', 'var')
+    error('Fatal error: Must run PCA_analysis before KNN_analysis; missing scores_table')
+end
+
+if ~exist('choice_variable', 'var')
+    error('Fatal error: Must run PCA_analysis before KNN_analysis; missing choice_variable')
+end
+
+%Extra required variables for this script
+if ~exist('make_median', 'var')
+    error('Fatal error: Missing required settings variable make_median in KNN_analysis')
+end
+
+if ~exist('make_full', 'var')
+    error('Fatal error: Missing required settings variable make_full in KNN_analysis')
+end
+  
+%% -- KNN analysis begins here -- 
+
+%% if doing full analysis with randomized labels (most likely just for backwads_select purposes)
+
+
+%% Using each image after PCA
+% using scores table--reassigning to new variable here to make it flexible
+if make_full
+    
+    knn_type = 'full';
+    
+    knn_with_apply = false;
+    
+    knn_table = scores_table;
+        
+    knn_data = knn_table{:,numeric_final_data_cols};
+
+    knn_drugs = knn_table.(choice_extension);
+
+    knn_helper
+end 
+%% Using medians after PCA
+
+if make_median 
+    tit_add = "";
+
+    knn_type = 'median';
+    knn_with_apply = false;
+    if after_pca
+        % if after pca, want to use scores table
+        knn_after_pca = "KNN after PCA";
+        % use separate script to get median values for each set of drugs in
+        % knn_table 
+        [drug_medians,drugs] = table_median(scores_table,choice_extension);
+    else
+        % if before pca, want to use pca_whitened_table
+        knn_after_pca = "KNN before PCA";
+        % use separate script to get median values for each set of drugs in
+        % knn_table 
+        [drug_medians,drugs] = table_median(pca_whitened_table,choice_extension);
+    end
+  
+    
+    %~~~ Perform knn ~~~%
+    % go through each drug, make a separate row for the indiviudal drug and
+    % compare it against the remainder 
+
+    % use medians as knn_data
+    knn_data = drug_medians;
+    knn_drugs = drugs';
+
+    disp("list of drugs for knn in knn_analysis")
+    disp(knn_drugs)
+    % run helper script 
+    knn_helper
+
+    % compare to random labels 
+    if random_compare
+        disp("Rerunning with random labels")
+        tit_add = " with random labels";
+        % copy list of knn_drusg 
+        random_drugs = randomize_list(knn_drugs);
+        % assign knn_drugs to be the new randomly created set
+        knn_drugs = random_drugs;
+        knn_helper 
+    end 
+end
+
+
+%% knn for applied -> individual
+
+if apply
+    if make_full
+        knn_type = 'full';
+        knn_with_apply = true;
+        % using scores table--reassigning to new variable here to make it flexible
+        knn_table = newspace_table;
+
+        knn_data = knn_table{:,numeric_final_data_cols};
+
+        knn_drugs = knn_table.DRUG;
+
+        knn_helper
+    end 
+end 
+
+%% create knn for applied 
+
+if apply
+    if make_median
+        knn_type = 'median';
+        knn_with_apply = true;
+
+        % get median values with external function 
+        [drug_medians,drugs] = table_median(newspace_table,choice_extension);
+        
+    %% Perform knn
+        % go through each drug, make a separate row for the indiviudal drug and
+        % compare it against the remainder 
+
+        % use medians as knn_data
+        knn_data = drug_medians;
+
+        knn_drugs = drugs';
+
+        knn_helper 
+        
+        disp("did applied knn")
+    end  
+end 
\ No newline at end of file

classification_and_analysis/analysis/backwards_select_analysis.m

+%% backwards selecvt analysis
+% all of the parts of backwards select without the prompts to load or save
+
+% set this as as safety 
+if ~exist('apply','var')
+    apply = false;
+end 
+
+if ~exist('no_TVN','var')
+    no_TVN = false; 
+end 
+
+%% Start backwards select 
+% start off by getting original percet 
+disp(strcat(num2str(pct_correct), "% for ", num2str(numvar), " variables"))
+
+% this is what we are looking to beat
+baseline_score = pct_correct;
+
+% minumum variable number to run until 
+minvar = 22; 
+
+% set keep_going to true to get while loop started
+keep_going = true;
+
+% save all of the best variable iterations
+overall_vars = struct; 
+
+% use external function to get git info
+git_info = getGitInfo();
+
+% can save hash info in overall vars
+try
+    overall_vars.hash =  git_info.hash; 
+catch
+    % have a catch in case repository wasn't cloned 
+    disp("unable to to save git hash information")
+end 
+
+% store the drugs used to create this in overall_vars 
+overall_vars.DRUGS = unique(final_data_table.DRUG)';
+% hard coding this for now b/c we're having problems 
+% remove_after_TVN = true;
+if remove_after_TVN
+    % not actually using untreated, want to remove from .DRUG for clarity
+    all_drugs_from_final = unique(final_data_table.DRUG)';
+    unt_loca = find(strcmp(all_drugs_from_final,"Untreated"));
+    all_drugs_from_final(unt_loca) = [];
+    
+    overall_vars.DRUGS = all_drugs_from_final;
+end 
+% save the workspace used
+overall_vars.WORKSPACE = chosen_workspaces;
+
+% save whether efflux was in lipid or not
+overall_vars.EFFLUX_IN_LIPID = efflux_in_lipid;
+
+
+disp(strcat("distance metric is ", d_metric))
+
+if large_group
+    disp("using broader groups")
+else
+    disp("using finer groups")
+end
+
+
+% use randomize_labels function on the drug column in final_data_table
+if with_randomized_labels
+    
+    % determine the randomized order
+    rand_order = randperm(length(final_data_table.DRUG));
+
+    disp("random!")
+    overall_vars.RANDOM = true;
+end 
+
+% store in the overall_vars if joint profile or not 
+if do_joint_profile
+    overall_vars.JOINT_PROFILE = true;
+else
+    overall_vars.JOINT_PROFILE = false;
+end 
+
+% once the max value is not going up anymore, we stop this loop
+disp("Beginning backwards select....")
+disp("value of remove_after_TVN")
+disp(remove_after_TVN)
+
+while keep_going
+    
+    % want to make an empty % error thing to store
+    all_pcts = zeros(length(starting_vars),1);
+    
+    % reset numvar 
+    numvar = length(kept_vars);
+    
+    % get title for structure
+    field_title = strcat("vars_",num2str(numvar),"_pct_",num2str(floor(baseline_score)));
+    
+    % save current number of variables in structure 
+    overall_vars.(field_title) = kept_vars;
+    
+    %%% might also want to create a confusion matrix and save its handle in
+    %%% overall_vars? 
+    
+    %% Now want to loop through, removing one variable at a time from starting_vars and calculating error 
+    for p = 1:length(kept_vars)
+        % reset to the full list of variables
+        starting_vars = kept_vars;
+        % get the name of the var removed
+        var_removed = starting_vars{p};
+
+        % remove a variable from the list
+        starting_vars(p) = [];
+        
+        % perform TVN with newly selected features
+        if no_TVN
+            pca_whitened_table = final_data_table; 
+        else
+            TVN_transform
+        end 
+        % if applied drug, remove here
+        if apply
+            
+            for qi = addedDrug
+                added_drug_1 = qi{1};
+                applied_inds = find(strcmp(pca_whitened_table.(choice_extension),added_drug_1));
+            % remove!!!
+                pca_whitened_table(applied_inds,:) = [];
+            end 
+ 
+        end 
+        
+        % if we want to get rid of the untreated we gotta do it every time
+        %% Remove after TVN
+        % forcing this true for now but change later
+        if remove_after_TVN
+            drugs = unique(pca_whitened_table.DRUG)';
+
+            drug_indexes = cell2struct(cell(1,length(drugs)), drugs, 2);
+
+            for i = drugs
+                drug = i{1};
+                drug_indexes.(drug) = find(strcmp(pca_whitened_table.DRUG, drug)).';
+            end
+
+            pca_whitened_table(drug_indexes.Untreated,:) = [];
+
+            drugs = unique(pca_whitened_table.DRUG)';
+
+            drug_indexes = cell2struct(cell(1,length(drugs)), drugs, 2);
+
+            for i = drugs
+                drug = i{1};
+                drug_indexes.(drug) = find(strcmp(pca_whitened_table.DRUG, drug)).';
+            end
+
+        end 
+        %% with randomized labels 
+        if with_randomized_labels
+            % go through each drug and reorder in the random order defined 
+            randomized_drugs = pca_whitened_table.DRUG; 
+            for w = 1:length(randomized_drugs)
+                randomized_drugs(w) = pca_whitened_table.DRUG(rand_order(w));
+            end 
+
+            pca_whitened_table.DRUG = randomized_drugs;
+
+        end 
+        %% now time for the analysis 
+        %find numeric columns of final table
+        numeric_final_data_cols = varfun(@isnumeric,pca_whitened_table,'OutputFormat', 'uniform');
+
+        %find column names
+        numeric_final_col_names = pca_whitened_table.Properties.VariableNames(numeric_final_data_cols);
+        
+        if plot_before_tvn
+             %find numeric columns of final table
+            numeric_final_data_cols = varfun(@isnumeric,final_data_table,'OutputFormat', 'uniform');
+