ADSP PCA: Difference between revisions

From bradwiki
Jump to navigation Jump to search
(Created page with "The Alzheimer's Disease Sequencing Project ([https://www.niagads.org/adsp/content/home ADSP]) has made available WES data in vcf format. The following is a step-by-step analys...")
 
No edit summary
Line 1: Line 1:
The Alzheimer's Disease Sequencing Project ([https://www.niagads.org/adsp/content/home ADSP]) has made available WES data in vcf format. The following is a step-by-step analysis walkthrough of this dataset. Here the ultimate goal is to develop a platform for Alzheimer's Disease diagnosis based on genome sequencing information. The data for this analysis can be downloaded here:
t-Distributed Stochastic Neighbor Embedding (tSNE) is a technique like PCA that allows one perform dimensionality reduction for visualization purposes. Supposedly tSNE does better than PCA at revealing clusters in high-dimensional data. Whereas PCA only allows you to visualize two or three components directly against each at the same time -- tSNE uses math magic to coerce a high-dimensional dataset into either a 2D or 3D array.
 
: [https://drive.google.com/open?id=1e3tIbQhcDUF1vofAwf4oYVOl8XriC44v ADSP_WES_VCF_LATEST_RELEASE.mat].
 
Both the dataset and the following analysis code are in MATLAB format. Note however there is an [https://cran.r-project.org/web/packages/R.matlab/index.html R package] for reading .mat files. All MATLAB code and custom functions used in the demo walkthrough below can be downloaded here:
 
: [https://github.com/subroutines/genos GENOS GITHUB CODE REPO]


{{SmallBox|float=right|clear=none|margin=0px 0px 8px 18px|width=170px|font-size=13px|Other Analyses|txt-size=11px|
{{SmallBox|float=right|clear=none|margin=0px 0px 8px 18px|width=170px|font-size=13px|Other Analyses|txt-size=11px|
Line 14: Line 8:
5. [[ADSP Stats|Descriptive Statistics]]<br>
5. [[ADSP Stats|Descriptive Statistics]]<br>
}}
}}
t-SNE models each multi-dim object against a point on a euclidean surface in such a way that similar features are modeled by nearby point functions and dissimilar features are modeled by distant point functions. It then projects these points onto the plane allowing you visualize, what would effectively be, all the interesting principal component combinations - the ones that yield unique clusters - simultaneously.
Again, this code can be downloaded from the: [https://github.com/subroutines/genos GENOS GIT]
<br> <br> <br>
==t-SNE Code==
----
<br><br>
<syntaxhighlight lang="matlab" line start="1" highlight="1" enclose="div">
% ######################################################################
%%      tSNE : t-Distributed Stochastic Neighbor Embedding
% ######################################################################
clc; close all; clear; rng('shuffle')
cd(fileparts(which('GENOS.m')));
MATDATA = 'ADSPdata.mat';
which(MATDATA)
load(MATDATA)
clearvars -except ADSP
%% CARBON COPY MAIN VARIABLES FROM ADSP.STRUCT
LOCI = ADSP.LOCI(:,1:17);
CASE = ADSP.CASE;
CTRL = ADSP.CTRL;
PHEN = ADSP.PHEN;
clearvars -except ADSP LOCI CASE CTRL PHEN
%###############################################################
%%      DETERMINE WHICH PARTICIPANTS TO KEEP
%###############################################################
PHE = PHEN(PHEN.TOTvars>14000,:);
PHECASE = PHE(PHE.AD==1,:);
PHECTRL = PHE(PHE.AD==0,:);
clearvars -except ADSP LOCI CASE CTRL PHEN PHE PHECASE PHECTRL
%###############################################################
%%          COUNT NUMBER OF VARIANTS PER LOCI
%###############################################################
% The varsum() function will go through each known variant loci
% and check whether anyone's SRR ID from your subset of IDs match
% all known SRR IDs for that loci. It will then sum the total
% number of alleles (+1 for hetzy-alt, +2 for homzy-alt) for each
% loci and return the totals.
[CASEN, CTRLN] = varsum(CASE, PHECASE.SRR, CTRL, PHECTRL.SRR);
% SAVE COUNTS AS NEW TABLE COLUMNS
LOCI.CASEREFS = numel(PHECASE.SRR)*2-CASEN;
LOCI.CTRLREFS = numel(PHECTRL.SRR)*2-CTRLN;
LOCI.CASEALTS = CASEN;
LOCI.CTRLALTS = CTRLN;
clearvars -except ADSP LOCI CASE CTRL PHEN PHE PHECASE PHECTRL
%###############################################################
%%              COMPUTE FISHER'S P-VALUE
%###############################################################
% COMPUTE FISHERS STATISTICS FOR THE TRAINING GROUP
[FISHP, FISHOR] = fishp_mex(LOCI.CASEREFS,LOCI.CASEALTS,...
                            LOCI.CTRLREFS,LOCI.CTRLALTS);
LOCI.FISHPS  = FISHP;
LOCI.FISHORS = FISHOR;
clearvars -except ADSP LOCI CASE CTRL PHEN PHE PHECASE PHECTRL
%% MAKE LATEST COUNTS THE MAIN TABLE STATS
LOCI.CASEREF = LOCI.CASEREFS;
LOCI.CTRLREF = LOCI.CTRLREFS;
LOCI.CASEALT = LOCI.CASEALTS;
LOCI.CTRLALT = LOCI.CTRLALTS;
LOCI.FISHP  = LOCI.FISHPS;
LOCI.FISHOR  = LOCI.FISHORS;
%% SORT VARIANT LOCI TABLE BY FISHER P-VALUE
[X,i] = sort(LOCI.FISHP);
LOCI  = LOCI(i,:);
CASE  = CASE(i);
CTRL  = CTRL(i);
LOCI.VID = (1:size(LOCI,1))';
LOCI.GENE = string(LOCI.GENE);
clc; clearvars -except ADSP LOCI CASE CTRL PHEN PHE PHECASE PHECTRL
disp(LOCI(1:9,:))
%% STORE VARIABLES FOR PCA/TSNE AS 'AMX'
AMX        = LOCI;
AMXCASE    = CASE;
AMXCTRL    = CTRL;
clearvars -except ADSP LOCI CASE CTRL PHEN PHE PHECASE PHECTRL...
AMX AMXCASE AMXCTRL
%% FILTER VARIANTS BASED ALT > REF
PASS = (AMX.CASEREF > AMX.CASEALT./1.5) | (AMX.CTRLREF > AMX.CTRLALT./1.5);
sum(~PASS)
AMX      = AMX(PASS,:);
AMXCASE  = AMXCASE(PASS);
AMXCTRL  = AMXCTRL(PASS);
AMX.VID  = (1:size(AMX,1))';
clearvars -except ADSP LOCI CASE CTRL PHEN PHE PHECASE PHECTRL...
AMX AMXCASE AMXCTRL
%% TAKE THE TOP N NUMBER OF VARIANTS
N = 100;
AMX      = AMX(1:N,:);
AMXCASE  = AMXCASE(1:N);
AMXCTRL  = AMXCTRL(1:N);
AMX.VID  = (1:size(AMX,1))';
fprintf('\n %.0f final loci count \n\n',size(AMX,1))
clearvars -except ADSP LOCI CASE CTRL PHEN PHE PHECASE PHECTRL...
AMX AMXCASE AMXCTRL
%% MAKE  RECTANGLE  NN VARIANT MATRIX
[ADNN, caMX, coMX] = varmx(AMX,AMXCASE,AMXCTRL,PHE);
clearvars -except ADSP LOCI CASE CTRL PHEN PHE PHECASE PHECTRL...
AMX AMXCASE AMXCTRL ADNN
%% RANDOMIZE ADNN AND REORDER PHE TO MATCH ADNN
ADL = ADNN(1,:);
ADN = ADNN(2:end,:);
i = randperm(size(ADN,1));
ADN = ADN(i,:);
ADNN = [ADL;ADN];
[i,j] = ismember(PHE.SRR, ADN(:,1) );
PHE.USED = i;
PHE.ORDER = j;
PHE = PHE(PHE.USED,:);
PHE = sortrows(PHE,'ORDER');
PCAMX = ADNN(2:end,4:end);
clearvars -except ADSP LOCI CASE CTRL PHEN PHE PHECASE PHECTRL...
AMX AMXCASE AMXCTRL ADNN PCAMX 
%% (OPTIONAL) PRE-PERFORM PCA BEFORE TSNE
% ss = statset('pca');
% ss.Display = 'iter';
% ss.MaxIter = 100;
% ss.TolFun = 1e4;
% ss.TolX = 1e4;
% ss.UseParallel = true;
%
% [PCAC,PCAS,~,~,~] = pca(  PCAMX' , 'Options',ss);
% clc; close all; scatter(PCAC(:,1),PCAC(:,2))
%
% % ...,'NumPCAComponents',0,...  means don't use PCA
% tSN = tsne(PCAC(:,1:10),'NumDimensions',2,'Theta',.6,'NumPCAComponents',0);
%
% clearvars -except ADSP GENB LOCI CASE CTRL PHEN AMX AMXCASE AMXCTRL...
% PHE ADNN PCAMX tSN PCAC PCAS
%######################################################################
%%      tSNE : t-Distributed Stochastic Neighbor Embedding
%######################################################################
tSN = tsne(PCAMX,'NumDimensions',2,'Theta',.6,'NumPCAComponents',8);
disp('done')
clearvars -except ADSP GENB LOCI CASE CTRL PHEN AMX AMXCASE AMXCTRL...
PHE ADNN PCAMX tSN PCAC PCAS
</syntaxhighlight>
==t-SNE Plots==
----
<br><br>
====ALZHEIMER'S STATUS====
<syntaxhighlight lang="matlab" line start="1" highlight="1" enclose="div">
%% PLOT TSNE --- ALZHEIMER'S STATUS (CASE/CTRL) --------------------------
close all;
fh1=figure('Units','normalized','Position',[.05 .05 .70 .84],'Color','w');
ax1=axes('Position',[.05 .02 .9 .9],'Color','none');
ph1 = gscatter(tSN(:,1),tSN(:,2),  PHE.AD, [],'.',15);
title({'\fontsize{16} t-SNE : CASE vs CTRL',' '})
legend(ph1,{'CTRL','CASE'},'FontSize',12,'Box','off','Location','NorthWest');
axis off
</syntaxhighlight>
<big>Top 100 variants</big>
[[File: TSNE Case Control.png|800px]]
<big>Top 2000 variants</big>
[[File: TSNE Case Control 2kvars.png|800px]]
====STUDY COHORT====
<syntaxhighlight lang="matlab" line start="1" highlight="1" enclose="div">
%% PLOT TSNE --- CONSORTIUM STUDY COHORT (1:24) -------------------------
close all;
fh1=figure('Units','normalized','Position',[.05 .05 .70 .84],'Color','w');
ax1=axes('Position',[.05 .02 .9 .9],'Color','none');
ph1 = gscatter(tSN(:,1),tSN(:,2),  PHE.COHORT, [],'.',15);
title({'\fontsize{16} t-SNE : STUDY COHORT',' '})
% legend(ph1,{'CTRL','CASE'},'FontSize',12,'Box','off','Location','NorthWest');
axis off
</syntaxhighlight>
<big>Top 100 variants</big>
[[File: TSNE Study Cohort.png|800px]]
<big>Top 2000 variants</big>
[[File: TSNE Study Cohort 2kvars.png|800px]]
====SEX====
<syntaxhighlight lang="matlab" line start="1" highlight="1" enclose="div">
%% PLOT TSNE --- SEX (M/F) ----------------------------------------------
close all;
fh1=figure('Units','normalized','Position',[.05 .05 .70 .84],'Color','w');
ax1=axes('Position',[.05 .02 .9 .9],'Color','none');
ph1 = gscatter(tSN(:,1),tSN(:,2),  PHE.SEX, [],'.',15);
title({'\fontsize{16} t-SNE : SEX',' '})
legend(ph1,{'Male','Female'},'FontSize',12,'Box','off','Location','NorthWest');
axis off
</syntaxhighlight>
<big>Top 100 variants</big>
[[File: TSNE Sex.png|800px]]
<big>Top 2000 variants</big>
[[File: TSNE Sex 2kvars.png|800px]]
====AGE====
<syntaxhighlight lang="matlab" line start="1" highlight="1" enclose="div">
%% PLOT TSNE --- AGE (BINNED AGE) ---------------------------------------
close all;
fh1=figure('Units','normalized','Position',[.05 .05 .70 .84],'Color','w');
ax1=axes('Position',[.05 .02 .9 .9],'Color','none');
AGE = round(PHE.AGE);
ofAGE = AGE>60;
A = AGE(ofAGE);
histogram(AGE)
[Y,E] = discretize(A,[60 80 90 91]);
% [Y,E] = discretize(A,[60 75 85 90 91]);
for nn = 1:numel(E)
A(Y==nn) = E(nn);
end
ph1 = gscatter(tSN(ofAGE,1),tSN(ofAGE,2),  A, [],'.',15);
title({'\fontsize{16} t-SNE : AGE',' '})
% legend(ph1,{'CTRL','CASE'},'FontSize',12,'Box','off','Location','NorthWest');
axis off
</syntaxhighlight>
<big>Top 100 variants</big>
[[File: TSNE Age.png|800px]]
<big>Top 2000 variants</big>
[[File: TSNE Age 2kvars.png|800px]]
====APOE STATUS====
<syntaxhighlight lang="matlab" line start="1" highlight="1" enclose="div">
%% PLOT TSNE --- APOE STATUS (22,23,24,33,34,44) ------------------------
close all;
fh1=figure('Units','normalized','Position',[.05 .05 .70 .84],'Color','w');
ax1=axes('Position',[.05 .02 .9 .9],'Color','none');
ph1 = gscatter(tSN(:,1),tSN(:,2),  PHE.APOE, [],'.',15);
ph1(1).MarkerSize = 35;
ph1(2).MarkerSize = 25;
ph1(2).Color = [.20 .20 .99];
ph1(3).MarkerSize = 35;
ph1(4).Color = [.99 .50 .10];
ph1(5).Color = [.30 .70 .80];
ph1(6).MarkerSize = 25;
title({'\fontsize{16} t-SNE : APOE',' '})
% legend(ph1,{'CTRL','CASE'},'FontSize',12,'Box','off','Location','NorthWest');
axis off
</syntaxhighlight>
<big>Top 100 variants</big>
[[File: TSNE APOE.png|800px]]
<big>Top 2000 variants</big>
[[File: TSNE APOE 2kvars.png|800px]]
====CONSENT GROUP====
<syntaxhighlight lang="matlab" line start="1" highlight="1" enclose="div">
%% PLOT TSNE --- CONSENT GROUP ------------------------------------------
close all;
fh1=figure('Units','normalized','Position',[.05 .05 .70 .84],'Color','w');
ax1=axes('Position',[.05 .02 .9 .9],'Color','none');
ph1 = gscatter(tSN(:,1),tSN(:,2),  PHE.RD, [],'.',15);
title({'\fontsize{16} t-SNE : CONSENT GROUP',' '})
% legend(ph1,{'CTRL','CASE'},'FontSize',12,'Box','off','Location','NorthWest');
axis off
</syntaxhighlight>
<big>Top 2000 variants</big>
[[File: TSNE Consent 2kvars.png|800px]]








<br> <br> <br> <br> <br> <br> <br>


==Additional Genomics Analyses==
==Additional Genomics Analyses==

Revision as of 05:22, 6 February 2018

t-Distributed Stochastic Neighbor Embedding (tSNE) is a technique like PCA that allows one perform dimensionality reduction for visualization purposes. Supposedly tSNE does better than PCA at revealing clusters in high-dimensional data. Whereas PCA only allows you to visualize two or three components directly against each at the same time -- tSNE uses math magic to coerce a high-dimensional dataset into either a 2D or 3D array.

Other Analyses


t-SNE models each multi-dim object against a point on a euclidean surface in such a way that similar features are modeled by nearby point functions and dissimilar features are modeled by distant point functions. It then projects these points onto the plane allowing you visualize, what would effectively be, all the interesting principal component combinations - the ones that yield unique clusters - simultaneously.

Again, this code can be downloaded from the: GENOS GIT




t-SNE Code



% ######################################################################
%%       tSNE : t-Distributed Stochastic Neighbor Embedding
% ######################################################################
clc; close all; clear; rng('shuffle')
cd(fileparts(which('GENOS.m')));


MATDATA = 'ADSPdata.mat';
which(MATDATA)
load(MATDATA)

clearvars -except ADSP



%% CARBON COPY MAIN VARIABLES FROM ADSP.STRUCT

LOCI = ADSP.LOCI(:,1:17);
CASE = ADSP.CASE;
CTRL = ADSP.CTRL;
PHEN = ADSP.PHEN;

clearvars -except ADSP LOCI CASE CTRL PHEN





%###############################################################
%%       DETERMINE WHICH PARTICIPANTS TO KEEP
%###############################################################



PHE = PHEN(PHEN.TOTvars>14000,:);


PHECASE = PHE(PHE.AD==1,:);
PHECTRL = PHE(PHE.AD==0,:);


clearvars -except ADSP LOCI CASE CTRL PHEN PHE PHECASE PHECTRL









%###############################################################
%%          COUNT NUMBER OF VARIANTS PER LOCI
%###############################################################

% The varsum() function will go through each known variant loci
% and check whether anyone's SRR ID from your subset of IDs match
% all known SRR IDs for that loci. It will then sum the total
% number of alleles (+1 for hetzy-alt, +2 for homzy-alt) for each
% loci and return the totals.


[CASEN, CTRLN] = varsum(CASE, PHECASE.SRR, CTRL, PHECTRL.SRR);


% SAVE COUNTS AS NEW TABLE COLUMNS
LOCI.CASEREFS = numel(PHECASE.SRR)*2-CASEN;
LOCI.CTRLREFS = numel(PHECTRL.SRR)*2-CTRLN;
LOCI.CASEALTS = CASEN;
LOCI.CTRLALTS = CTRLN;


clearvars -except ADSP LOCI CASE CTRL PHEN PHE PHECASE PHECTRL







%###############################################################
%%               COMPUTE FISHER'S P-VALUE
%###############################################################


% COMPUTE FISHERS STATISTICS FOR THE TRAINING GROUP
[FISHP, FISHOR] = fishp_mex(LOCI.CASEREFS,LOCI.CASEALTS,...
                            LOCI.CTRLREFS,LOCI.CTRLALTS);

LOCI.FISHPS  = FISHP;
LOCI.FISHORS = FISHOR;


clearvars -except ADSP LOCI CASE CTRL PHEN PHE PHECASE PHECTRL





%% MAKE LATEST COUNTS THE MAIN TABLE STATS

LOCI.CASEREF = LOCI.CASEREFS;
LOCI.CTRLREF = LOCI.CTRLREFS;
LOCI.CASEALT = LOCI.CASEALTS;
LOCI.CTRLALT = LOCI.CTRLALTS;
LOCI.FISHP   = LOCI.FISHPS;
LOCI.FISHOR  = LOCI.FISHORS;






%% SORT VARIANT LOCI TABLE BY FISHER P-VALUE

[X,i] = sort(LOCI.FISHP);

LOCI  = LOCI(i,:);
CASE  = CASE(i);
CTRL  = CTRL(i);
LOCI.VID = (1:size(LOCI,1))';

LOCI.GENE = string(LOCI.GENE);



clc; clearvars -except ADSP LOCI CASE CTRL PHEN PHE PHECASE PHECTRL
disp(LOCI(1:9,:))





%% STORE VARIABLES FOR PCA/TSNE AS 'AMX'

AMX         = LOCI;
AMXCASE     = CASE;
AMXCTRL     = CTRL;


clearvars -except ADSP LOCI CASE CTRL PHEN PHE PHECASE PHECTRL...
AMX AMXCASE AMXCTRL 





%% FILTER VARIANTS BASED ALT > REF

PASS = (AMX.CASEREF > AMX.CASEALT./1.5) | (AMX.CTRLREF > AMX.CTRLALT./1.5);
sum(~PASS)

AMX      = AMX(PASS,:);
AMXCASE  = AMXCASE(PASS);
AMXCTRL  = AMXCTRL(PASS);
AMX.VID  = (1:size(AMX,1))';




clearvars -except ADSP LOCI CASE CTRL PHEN PHE PHECASE PHECTRL...
AMX AMXCASE AMXCTRL 





%% TAKE THE TOP N NUMBER OF VARIANTS
N = 100;


AMX      = AMX(1:N,:);
AMXCASE  = AMXCASE(1:N);
AMXCTRL  = AMXCTRL(1:N);
AMX.VID  = (1:size(AMX,1))';

fprintf('\n %.0f final loci count \n\n',size(AMX,1))

clearvars -except ADSP LOCI CASE CTRL PHEN PHE PHECASE PHECTRL...
AMX AMXCASE AMXCTRL 








%% MAKE  RECTANGLE  NN VARIANT MATRIX


[ADNN, caMX, coMX] = varmx(AMX,AMXCASE,AMXCTRL,PHE);

clearvars -except ADSP LOCI CASE CTRL PHEN PHE PHECASE PHECTRL...
AMX AMXCASE AMXCTRL ADNN 









%% RANDOMIZE ADNN AND REORDER PHE TO MATCH ADNN

ADL = ADNN(1,:);
ADN = ADNN(2:end,:);
i = randperm(size(ADN,1));
ADN = ADN(i,:);
ADNN = [ADL;ADN];


[i,j] = ismember(PHE.SRR, ADN(:,1) );
PHE.USED = i;
PHE.ORDER = j;
PHE = PHE(PHE.USED,:);
PHE = sortrows(PHE,'ORDER');



PCAMX = ADNN(2:end,4:end);


clearvars -except ADSP LOCI CASE CTRL PHEN PHE PHECASE PHECTRL...
AMX AMXCASE AMXCTRL ADNN PCAMX  




%% (OPTIONAL) PRE-PERFORM PCA BEFORE TSNE 

% ss = statset('pca');
% ss.Display = 'iter';
% ss.MaxIter = 100;
% ss.TolFun = 1e4;
% ss.TolX = 1e4;
% ss.UseParallel = true;
% 
% [PCAC,PCAS,~,~,~] = pca(  PCAMX' , 'Options',ss);
% clc; close all; scatter(PCAC(:,1),PCAC(:,2))
%
% % ...,'NumPCAComponents',0,...  means don't use PCA
% tSN = tsne(PCAC(:,1:10),'NumDimensions',2,'Theta',.6,'NumPCAComponents',0);
% 
% clearvars -except ADSP GENB LOCI CASE CTRL PHEN AMX AMXCASE AMXCTRL...
% PHE ADNN PCAMX tSN PCAC PCAS






%######################################################################
%%       tSNE : t-Distributed Stochastic Neighbor Embedding
%######################################################################



tSN = tsne(PCAMX,'NumDimensions',2,'Theta',.6,'NumPCAComponents',8);


disp('done')
clearvars -except ADSP GENB LOCI CASE CTRL PHEN AMX AMXCASE AMXCTRL...
PHE ADNN PCAMX tSN PCAC PCAS




t-SNE Plots





ALZHEIMER'S STATUS

%% PLOT TSNE --- ALZHEIMER'S STATUS (CASE/CTRL) --------------------------
close all; 
fh1=figure('Units','normalized','Position',[.05 .05 .70 .84],'Color','w');
ax1=axes('Position',[.05 .02 .9 .9],'Color','none');

ph1 = gscatter(tSN(:,1),tSN(:,2),  PHE.AD, [],'.',15);

title({'\fontsize{16} t-SNE : CASE vs CTRL',' '})
legend(ph1,{'CTRL','CASE'},'FontSize',12,'Box','off','Location','NorthWest');
axis off

Top 100 variants Error creating thumbnail: File missing

Top 2000 variants Error creating thumbnail: File missing




STUDY COHORT

%% PLOT TSNE --- CONSORTIUM STUDY COHORT (1:24) -------------------------
close all; 
fh1=figure('Units','normalized','Position',[.05 .05 .70 .84],'Color','w');
ax1=axes('Position',[.05 .02 .9 .9],'Color','none');

ph1 = gscatter(tSN(:,1),tSN(:,2),  PHE.COHORT, [],'.',15);


title({'\fontsize{16} t-SNE : STUDY COHORT',' '})
% legend(ph1,{'CTRL','CASE'},'FontSize',12,'Box','off','Location','NorthWest');
axis off

Top 100 variants Error creating thumbnail: File missing

Top 2000 variants Error creating thumbnail: File missing



SEX

%% PLOT TSNE --- SEX (M/F) ----------------------------------------------
close all; 
fh1=figure('Units','normalized','Position',[.05 .05 .70 .84],'Color','w');
ax1=axes('Position',[.05 .02 .9 .9],'Color','none');

ph1 = gscatter(tSN(:,1),tSN(:,2),  PHE.SEX, [],'.',15);


title({'\fontsize{16} t-SNE : SEX',' '})
legend(ph1,{'Male','Female'},'FontSize',12,'Box','off','Location','NorthWest');
axis off

Top 100 variants Error creating thumbnail: File missing

Top 2000 variants Error creating thumbnail: File missing



AGE

%% PLOT TSNE --- AGE (BINNED AGE) ---------------------------------------
close all; 
fh1=figure('Units','normalized','Position',[.05 .05 .70 .84],'Color','w');
ax1=axes('Position',[.05 .02 .9 .9],'Color','none');

AGE = round(PHE.AGE);
ofAGE = AGE>60;
A = AGE(ofAGE);

histogram(AGE)

[Y,E] = discretize(A,[60 80 90 91]);
% [Y,E] = discretize(A,[60 75 85 90 91]);
for nn = 1:numel(E)
A(Y==nn) = E(nn);
end

ph1 = gscatter(tSN(ofAGE,1),tSN(ofAGE,2),  A, [],'.',15);


title({'\fontsize{16} t-SNE : AGE',' '})
% legend(ph1,{'CTRL','CASE'},'FontSize',12,'Box','off','Location','NorthWest');
axis off

Top 100 variants Error creating thumbnail: File missing

Top 2000 variants Error creating thumbnail: File missing



APOE STATUS

%% PLOT TSNE --- APOE STATUS (22,23,24,33,34,44) ------------------------
close all; 
fh1=figure('Units','normalized','Position',[.05 .05 .70 .84],'Color','w');
ax1=axes('Position',[.05 .02 .9 .9],'Color','none');


ph1 = gscatter(tSN(:,1),tSN(:,2),  PHE.APOE, [],'.',15);


ph1(1).MarkerSize = 35;
ph1(2).MarkerSize = 25;
ph1(2).Color = [.20 .20 .99];
ph1(3).MarkerSize = 35;
ph1(4).Color = [.99 .50 .10];
ph1(5).Color = [.30 .70 .80];
ph1(6).MarkerSize = 25;

title({'\fontsize{16} t-SNE : APOE',' '})
% legend(ph1,{'CTRL','CASE'},'FontSize',12,'Box','off','Location','NorthWest');
axis off

Top 100 variants Error creating thumbnail: File missing

Top 2000 variants Error creating thumbnail: File missing



CONSENT GROUP

%% PLOT TSNE --- CONSENT GROUP ------------------------------------------
close all; 
fh1=figure('Units','normalized','Position',[.05 .05 .70 .84],'Color','w');
ax1=axes('Position',[.05 .02 .9 .9],'Color','none');


ph1 = gscatter(tSN(:,1),tSN(:,2),  PHE.RD, [],'.',15);


title({'\fontsize{16} t-SNE : CONSENT GROUP',' '})
% legend(ph1,{'CTRL','CASE'},'FontSize',12,'Box','off','Location','NorthWest');
axis off

Top 2000 variants Error creating thumbnail: File missing






















Additional Genomics Analyses



Other Analyses










Notes

Category:ADSP

Retrieved from "https://bradleymonk.com/wiki/index.php?title=ADSP_PCA&oldid=3824"