@@ -58,10 +58,9 @@ importFrom(mvtnorm, rmvnorm)

 importFrom(oligoClasses, celfileDate, chromosome2integer, i2p,

            initializeBigMatrix, initializeBigVector, integerMatrix,

-	   isPackageLoaded,

-           ldPath, ocLapply, ocProbesets, ocSamples,

-	   parStatus,

-           splitIndicesByLength, splitIndicesByNode, AssayDataList)

+           isPackageLoaded, ldPath, ocLapply, ocProbesets, ocSamples,

+           parStatus, splitIndicesByLength, splitIndicesByNode, 

+           AssayDataList, genomeBuild, celfileName)

 importFrom(preprocessCore, normalize.quantiles, normalize.quantiles.determine.target,

            normalize.quantiles.use.target, subColSummarizeMedian)

@@ -85,6 +84,8 @@ importFrom(parallel, makeCluster, detectCores, parRapply, stopCluster, clusterEv

 importFrom(limma, loessFit)

+importFrom(ff, delete)

+

 ##----------------------------------------------------------------------------

 ## export

 ##----------------------------------------------------------------------------

@@ -25,8 +25,7 @@ krlmm <- function(cnSet, cdfName, gender=NULL, trueCalls=NULL, verbose=TRUE) {

     colnames(scores) = colnames(M)  

     priormeans = calculatePriorValues(M, numSNP, verbose)

-    VGLMparameters = calculateParameters(M, priormeans, numSNP, verbose)

-  

+      

 ### retrieve or calculate coefficients

     krlmmCoefficients = getKrlmmVGLMCoefficients(pkgname, trueCalls, VGLMparameters, verbose, numSample, colnames(M))

@@ -83,7 +82,7 @@ krlmmnopkg <- function(cnSet, offset, gender=NULL, normalize.method=NULL, annota

     }

     message("Leaving out non-variant SNPs")

-

+    

     # For SNPs with less than 3 distinct data point, exclude them from downstream analysis

     uniqueCount = apply(M[,], 1, function(x){length(unique(x))})

     SNPtoProcessIndex = uniqueCount >= 3

@@ -113,8 +112,18 @@ krlmmnopkg <- function(cnSet, offset, gender=NULL, normalize.method=NULL, annota

     }

     colnames(trueCalls)=colnames(M)

     rownames(trueCalls)=rownames(M)

-

-    priormeans = calculatePriorValues(M, numSNP, verbose)

+if(numSample<=30){

+  priormeans=prepriormeans

+}else{

+  priormeans = calculatePriorValues(M, numSNP, verbose)

+    if((abs(priormeans[1]-prepriormeans[1])+abs(priormeans[2]-prepriormeans[2])+abs(priormeans[3]-prepriormeans[3]))>=3){

+priormeans=prepriormeans

+}else{

+      priormeans=priormeans

+  }

+ }

+ 

+    

     VGLMparameters = calculateParameters(M, priormeans, numSNP, verbose)

 ### retrieve or calculate coefficients

@@ -131,23 +140,24 @@ krlmmnopkg <- function(cnSet, offset, gender=NULL, normalize.method=NULL, annota

 ### assign confidence scores

     computeCallPr(scores, M, calls, numSNP, numSample, verbose)

-    YIndex = seq(1:nrow(cnSet))[chromosome(cnSet)==24 & isSnp(cnSet)]

-

+   YIndex = seq(1:nrow(cnSet))[chromosome(cnSet)==24 & isSnp(cnSet)]

+    XIndex = seq(1:nrow(cnSet))[chromosome(cnSet)==23 & isSnp(cnSet)]

 ### impute gender if gender information not provided

     if (is.null(gender)) {

-        gender = krlmmImputeGender(cnSet, annotation, YIndex, verbose, offset=offset)

+       gender = krlmmImputeGender(cnSet, annotation, YIndex, verbose, offset=offset)

     }

-### double-check ChrY SNP cluster, impute gender if gender information not provided

+### double-check ChrY ChrX SNP cluster, impute gender if gender information not provided

     if (!(is.null(gender))) {

         verifyChrYSNPs(cnSet, M, calls, gender, annotation, YIndex, priormeans, verbose)

+        verifyChrXSNPs(cnSet, M, calls, gender, annotation, XIndex, priormeans, verbose)

     }

 #    if (length(YIndex)>0  && !(is.null(gender))) {

 #      verifyChrYSNPs(cnSet, M, calls, gender, annotation, YIndex, priormeans, verbose)

 #    }

 ### add back SNPs excluded before

-    AddBackNoVarianceSNPs(cnSet, calls, scores, numSNP, numSample, SNPtoProcessIndex, noVariantSNPIndex)

+   AddBackNoVarianceSNPs(cnSet, calls, scores, numSNP, numSample, SNPtoProcessIndex, noVariantSNPIndex)

     close(calls)

     close(scores)

@@ -170,7 +180,7 @@ loessnormalization<- function(cnSet, verbose, offset=16, blockSize = 300000){

     numSNP <- nrow(A)

     numSample <- ncol(A)

-    library(limma)

+   

     M <- oligoClasses::initializeBigMatrix(name="M", numSNP, numSample, vmode = "double")

     S <- oligoClasses::initializeBigMatrix(name="S", numSNP, numSample, vmode = "double")

@@ -927,7 +937,7 @@ krlmmImputeGender <- function(cnSet, annotation, YIndex, verbose, offset=0){

     S = computeAverageLogIntensity(cnSet, verbose, offset=offset) 

     # S is calculated and saved in original SNP order. 

-    matchy = match(annotation[YIndex, "Name"], rownames(S))

+    matchy = match(rownames(annotation)[YIndex], rownames(S))

     matchy = matchy[!is.na(matchy)]

     if (length(matchy) <= 10){

         predictedGender = rep(NA, ncol(S))

@@ -955,8 +965,8 @@ krlmmImputeGender <- function(cnSet, annotation, YIndex, verbose, offset=0){

     priorS = apply(allS, 2, FUN="median", na.rm=TRUE)

     if (abs(priorS[1] - priorS[2]) <= 1.6) {

-        message("Separation between clusters too small (samples probabaly all the same gender): skipping gender prediction step");

-        predictedGender = rep(NA, ncol(Sy))

+        message("Separation between clusters too small (samples probabaly all the same gender): ");

+      

     }

     meanmatrix = apply(Sy, 2, median)

@@ -986,7 +996,7 @@ verifyChrYSNPs <- function(cnSet, M, calls, gender, annotation, YIndex, priormea

     open(callsGt)

     open(callsPr)

-    matchy = match(annotation[YIndex, "Name"], rownames(M))

+    matchy = match(rownames(annotation)[YIndex], rownames(M))

     matchy = matchy[!is.na(matchy)]

     MChrY = M[matchy,]

@@ -1031,3 +1041,58 @@ verifyChrYSNPs <- function(cnSet, M, calls, gender, annotation, YIndex, priormea

     if (verbose) message("Done verifying SNPs on Chromosome Y")

 }

+

+

+

+

+#######################################

+verifyChrXSNPs <- function(cnSet, M, calls, gender, annotation, XIndex, priormeans, verbose){

+    if (verbose) message("Start verifying SNPs on Chromosome X")

+    callsGt = calls(cnSet)

+    callsPr = snpCallProbability(cnSet)

+    open(callsGt)

+    open(callsPr)

+      

+    matchx = match(rownames(annotation)[XIndex], rownames(M))

+    matchx = matchx[!is.na(matchx)]

+   

+    MChrX= M[matchx,]

+    callsChrX = callsGt[matchx,]

+  

+    male = gender == 1

+    female = gender == 2    

+    

+    checkK = apply(callsChrX[, male], 1, function(x){ length(unique(x[!is.na(x)])) } )

+    

+    for(i in 1:nrow(MChrX)){

+      

+        # Chromosome X SNPs for male, k = 1 or 2 permitted for male samples

+thisChrXSNPorigPosition = match(rownames(callsChrX)[i], rownames(callsGt))

+        # early exit for k = 1 or 2 cases. Only re-assign calls to male samples if we previouly assigned

+        # male samples to 3 clusters by mistake. 

+        if (checkK[i] < 3) next;

+          

+        if (class(try(kmeans(MChrX[i, male], c(priormeans[1], priormeans[3]), nstart=1), TRUE)) != "try-error"){

+           

+            maleSampleCalls = kmeans(MChrX[i, male],c(priormeans[1], priormeans[3]), nstart=45)$cluster

+            callsGt[thisChrXSNPorigPosition, male][maleSampleCalls == 1] = 1

+            callsGt[thisChrXSNPorigPosition, male][maleSampleCalls == 2] = 3

+        } else {

+                    

+            distanceToPrior1 = mean(abs(MChrX[i, male] - priormeans[1]))

+            distanceToPrior3 = mean(abs(MChrX[i, male] - priormeans[3]))                

+            callsGt[thisChrXSNPorigPosition, male] = ifelse(distanceToPrior1 < distanceToPrior3, 1, 3)

+        }

+

+        MMaleSamples = MChrX[i, male]

+        callsMaleSample = callsGt[thisChrXSNPorigPosition, male]

+        scoresMaleSample = .Call("krlmmConfidenceScore", t(as.matrix(MMaleSamples)), t(as.matrix(callsMaleSample)), PACKAGE="crlmm");

+

+       callsPr[thisChrXSNPorigPosition, male] = scoresMaleSample       

+    }

+

+    close(callsGt)

+    close(callsPr)

+    

+    if (verbose) message("Done verifying SNPs on Chromosome X")

+}

@@ -422,12 +422,12 @@ SEXP krlmmComputeM(SEXP A, SEXP B){

     INTERNAL VARIABLES

     -------- ---------

-    rowsAB, colsAB: dimensions of A and B, which is number of SNP and number of sample

+    num_SNP, num_sample: dimensions of A and B, which is number of SNP and number of sample

   */

-  long rowsAB, colsAB;

-  rowsAB = INTEGER(getAttrib(A, R_DimSymbol))[0];

-  colsAB = INTEGER(getAttrib(A, R_DimSymbol))[1];

+  long num_SNP, num_sample;

+  num_SNP = INTEGER(getAttrib(A, R_DimSymbol))[0];

+  num_sample = INTEGER(getAttrib(A, R_DimSymbol))[1];

   A = coerceVector(A, REALSXP);

   B = coerceVector(B, REALSXP);

@@ -435,16 +435,16 @@ SEXP krlmmComputeM(SEXP A, SEXP B){

   long i, j;

   SEXP Rval;

-  PROTECT(Rval = allocMatrix(REALSXP, rowsAB, colsAB));

+  PROTECT(Rval = allocMatrix(REALSXP, num_SNP, num_sample));

   double *ptr2M;

   long elepos;

   ptr2M = NUMERIC_POINTER(Rval);

-  for (i = 1; i <= rowsAB; i++){

-    for (j = 1; j <= colsAB; j++){

+  for (i = 1; i <= num_SNP; i++){

+    for (j = 1; j <= num_sample; j++){

       // elepos is an index for A, B and M

-      elepos = CMatrixElementPosition(i, j, rowsAB);

+      elepos = CMatrixElementPosition(i, j, num_SNP);

       ptr2M[elepos] = (log2(REAL(A)[elepos])-log2(REAL(B)[elepos]));

     }

   }

@@ -459,34 +459,34 @@ SEXP krlmmComputeS(SEXP A, SEXP B){

   /*

     ARGUMENTS

     ---------

-    A: intensity matrix for allele A11

+    A: intensity matrix for allele A

     B: intensity matrix for allele B

     S: average log-intensity for a particular SNP (outgoing)

     INTERNAL VARIABLES

     -------- ---------

-    rowsAB, colsAB: dimensions of A and B, which is number of SNP and number of sample

+    num_SNP, num_sample: dimensions of A and B, which is number of SNP and number of sample

   */

-  long rowsAB, colsAB;

-  rowsAB = INTEGER(getAttrib(A, R_DimSymbol))[0];

-  colsAB = INTEGER(getAttrib(A, R_DimSymbol))[1];

+  long num_SNP, num_sample;

+  num_SNP = INTEGER(getAttrib(A, R_DimSymbol))[0];

+  num_sample = INTEGER(getAttrib(A, R_DimSymbol))[1];

   A = coerceVector(A, REALSXP);

   B = coerceVector(B, REALSXP); 

   long i, j;

   SEXP Rval;

-  PROTECT(Rval = allocMatrix(REALSXP, rowsAB, colsAB));

+  PROTECT(Rval = allocMatrix(REALSXP, num_SNP, num_sample));

   double *ptr2S;

   long elepos;

   ptr2S = NUMERIC_POINTER(Rval);

-  for (i = 1; i <= rowsAB; i++){

-    for (j = 1; j <= colsAB; j++){

+  for (i = 1; i <= num_SNP; i++){

+    for (j = 1; j <= num_sample; j++){

       // elepos is an index for A, B and S

-      elepos = CMatrixElementPosition(i, j, rowsAB);

+      elepos = CMatrixElementPosition(i, j, num_SNP);

       ptr2S[elepos] = (log2(REAL(A)[elepos])+log2(REAL(B)[elepos]))/2;

     }

   }

@@ -495,15 +495,16 @@ SEXP krlmmComputeS(SEXP A, SEXP B){

   return Rval;

 }

-void calculate_multiple_cluster_scores(int row, double *intensity, double mean_intensity, int *clustering, int num_SNP, int num_sample, int *ptr, double **dist, int *clustercount){

-    int p, q, o;

+void calculate_multiple_cluster_scores(long row, double *intensity, double mean_intensity, int *clustering, long num_SNP, long num_sample, int *ptr, double **dist, long *clustercount){

+    long p, q;

+    int o;

     long vectorPos;

     double a_dist;

     for(p=1; p <= num_sample; p++){

         dist[p][p] = 0;

     }

-    for(p=1; p<num_sample; p++){

-        for(q=p+1; q<=num_sample; q++){

+    for(p=1; p < num_sample; p++){

+        for(q = p+1; q <= num_sample; q++){

    	    a_dist = fabs(intensity[CMatrixElementPosition(row, p, num_SNP)] - intensity[CMatrixElementPosition(row, q, num_SNP)]);

             dist[p][q] = a_dist;

             dist[q][p] = a_dist;

@@ -521,7 +522,7 @@ void calculate_multiple_cluster_scores(int row, double *intensity, double mean_i

         sum[2] = 0;

         sum[3] = 0;

         for (q=1; q<=num_sample; q++){

-	  sum[clustering[CMatrixElementPosition(row, q, num_SNP)]] += dist[p][q];

+            sum[clustering[CMatrixElementPosition(row, q, num_SNP)]] += dist[p][q];

         }

         within_cluster = sum[clustering[vectorPos]] / (clustercount[clustering[vectorPos]]- 1);

@@ -546,9 +547,9 @@ void calculate_multiple_cluster_scores(int row, double *intensity, double mean_i

     }   

 }

-void calculate_unique_cluster_scores(int row, double *intensity, double mean_intensity, double intensity_range, int num_SNP, int num_sample, int *ptr)				   

+void calculate_unique_cluster_scores(long row, double *intensity, double mean_intensity, double intensity_range, long num_SNP, long num_sample, int *ptr)				   

 {

-    int p;

+    long p;

     long vectorPos;

     for(p = 1; p <= num_sample; p++){

@@ -558,24 +559,24 @@ void calculate_unique_cluster_scores(int row, double *intensity, double mean_int

 }

-double calculate_SNP_mean(int row, double *intensity, int num_SNP, int num_sample)

+double calculate_SNP_mean(long row, double *intensity, long num_SNP, long num_sample)

 {

   double sum_intensity;

   sum_intensity = 0;

-  int p;

+  long p;

   for(p = 1; p <= num_sample; p++){

     sum_intensity = sum_intensity + intensity[CMatrixElementPosition(row, p, num_SNP)];

   }

   return(sum_intensity / num_sample);

 }

-double calculate_SNP_range(int row, double *intensity, int num_SNP, int num_sample)

+double calculate_SNP_range(long row, double *intensity, long num_SNP, long num_sample)

 {

   double min_intensity;

   double max_intensity;

   max_intensity = intensity[CMatrixElementPosition(row, 1, num_SNP)];

   min_intensity = intensity[CMatrixElementPosition(row, 1, num_SNP)];

-  int p;  

+  long p;  

   for(p = 2; p <= num_sample; p++){

     if (intensity[CMatrixElementPosition(row, p, num_SNP)] < min_intensity){

       min_intensity = intensity[CMatrixElementPosition(row, p, num_SNP)];  

@@ -590,7 +591,7 @@ double calculate_SNP_range(int row, double *intensity, int num_SNP, int num_samp

 SEXP krlmmConfidenceScore(SEXP M, SEXP clustering)

 {

-    int num_SNP, num_sample;

+    long num_SNP, num_sample;

     num_SNP = INTEGER(getAttrib(M, R_DimSymbol))[0];

     num_sample = INTEGER(getAttrib(M, R_DimSymbol))[1];

@@ -598,7 +599,7 @@ SEXP krlmmConfidenceScore(SEXP M, SEXP clustering)

     ptr2cluster = INTEGER_POINTER(AS_INTEGER(clustering));

     M = coerceVector(M, REALSXP);

-    int i, j;

+    long i, j;

     int cluster;

     double mean_intensity;

     double intensity_range;

@@ -617,34 +618,31 @@ SEXP krlmmConfidenceScore(SEXP M, SEXP clustering)

       dist[i] = (double *)malloc((num_sample + 1) * sizeof(double)); 

     }

-    int cluster_count[4]; // extra element to cope with zero-base notation

-   // int clid[4];

+    long cluster_count[4]; // extra element to cope with zero-base notation

     for (i = 1; i <= num_SNP; i++) {

         cluster_count[1] = 0;

         cluster_count[2] = 0;

         cluster_count[3] = 0;

         for (j=1; j<= num_sample; j++){

-	    cluster = ptr2cluster[CMatrixElementPosition(i, j, num_SNP)];

+            cluster = ptr2cluster[CMatrixElementPosition(i, j, num_SNP)];

             cluster_count[cluster]++;

         }

-	mean_intensity = calculate_SNP_mean(i, REAL(M), num_SNP, num_sample);

-	intensity_range = calculate_SNP_range(i, REAL(M), num_SNP, num_sample);

+        mean_intensity = calculate_SNP_mean(i, REAL(M), num_SNP, num_sample);

+        intensity_range = calculate_SNP_range(i, REAL(M), num_SNP, num_sample);

         k = 0;

         for (j=1; j<=3; j++){

             if (cluster_count[j] > 0){

                 k++;

-            //    clid[k] = j;

             }

         }

-	if (k==1) {

-	  calculate_unique_cluster_scores(i, REAL(M), mean_intensity, intensity_range, num_SNP, num_sample, ptr2score);

-	} else {

-	  calculate_multiple_cluster_scores(i, REAL(M), mean_intensity, ptr2cluster, num_SNP, num_sample, ptr2score, dist, cluster_count);

-	}       

-      

+        if (k==1) {

+            calculate_unique_cluster_scores(i, REAL(M), mean_intensity, intensity_range, num_SNP, num_sample, ptr2score);

+        } else {

+	        calculate_multiple_cluster_scores(i, REAL(M), mean_intensity, ptr2cluster, num_SNP, num_sample, ptr2score, dist, cluster_count);			

+        }       

     } 

     UNPROTECT(1);