int main(int argc, char **argv)
{
int **snppos ;
int *snpindx ;
char **snpnamelist, **indnamelist ;
char **eglist ;
int lsnplist, lindlist, numeg ;
int i,j;
SNP *cupt, *cupt1, *cupt2, *cupt3 ;
Indiv *indx ;
double gpos1,gpos2,cpos1,cpos2,gd, cd, gd100 ;
double rthresh, zt ;
int mpflag, ret, numvalidind, nvalid, numvalidsnps ;
int ch1, ch2 ;
int fmnum , lmnum ;
int num, n1, n2 ;
int nkill = 0 ;
int t, k, g ;
int nindiv = 0, e, f, lag=1 ;
double xc[9], xd[4], xc2[9] ;
double ychi, zscore, zthresh = 20.0 ;
double y1, y2 ;
int nignore, numrisks = 1 ;
char **genolist ;
int numgenolist ;
char c1, c2 ;
int t1, t2 ;
malexhet = YES ; // convertf default is don't change the data
tersem = YES ; // no snp counts
readcommands(argc, argv) ;
setomode(&outputmode, omode) ;
packmode = YES ;
settersemode(tersem) ;
if (r2thresh > 0.0) killr2 = YES ;
if (badpedignore) setbadpedignore() ;
numsnps =
getsnps(snpname, &snpmarkers, 0.0, badsnpname, &nignore, numrisks) ;
for (i=0; i<numsnps; i++) {
if (xchrom == -1) break ;
cupt = snpmarkers[i] ;
if (cupt -> chrom != xchrom) cupt -> ignore = YES ;
if (cupt -> ignore) continue ;
t = nnint(cupt -> physpos) ;
if ( (t< lopos) || (t >hipos)) cupt -> ignore = YES ;
}
nignore = 0 ;
for (i=0; i<numsnps; i++) {
cupt = snpmarkers[i] ;
if (cupt -> chrom > maxchrom) cupt -> ignore = YES ;
if (cupt -> chrom < minchrom) cupt -> ignore = YES ;
if (cupt -> ignore) ++nignore ;
}
if (numsnps == nignore) fatalx("no valid snps\n") ;
numindivs = getindivs(indivname, &indivmarkers) ;
if (polarid != NULL) {
polarindex = indindex(indivmarkers, numindivs, polarid) ;
if (polarindex<0) fatalx("polarid %s not found\n") ;
}
if (genotypelist!= NULL) {
getgenos_list(genotypelist, snpmarkers, indivmarkers,
numsnps, numindivs, nignore) ;
}
else {
setgenotypename(&genotypename, indivname) ;
getgenos(genotypename, snpmarkers, indivmarkers,
numsnps, numindivs, nignore) ;
}
if (newsnpname != NULL) {
nums2 =
getsnps(newsnpname, &snpm2, 0.0, NULL, &nignore, numrisks) ;
remap(snpmarkers, numsnps, snpm2, nums2) ;
snpmarkers = snpm2 ;
numsnps = nums2 ;
}
if (newindivname != NULL) {
numind2 = getindivs(newindivname, &indm2) ;
remapind(snpmarkers, numsnps, indivmarkers, indm2, numindivs, numind2) ;
indivmarkers = indm2 ;
numindivs = numind2 ;
if (polarid != NULL) {
polarindex = indindex(indivmarkers, numindivs, polarid) ;
}
}
if (mkdiploid) {
numindivs = rmindivs(snpmarkers, numsnps, indivmarkers, numindivs) ;
numind2 = mkindh2d(indivmarkers, &indm2, numindivs) ;
remaph2d(snpmarkers, numsnps, indivmarkers, indm2, numindivs, numind2) ;
indivmarkers = indm2 ;
numindivs = numind2 ;
}
if (deletedup) dedupit(snpmarkers, numsnps) ; // only one marker per position
for (i=0; i<numsnps; i++) {
cupt = snpmarkers[i] ;
if (zerodistance) cupt -> genpos = 0.0 ;
c1 = cupt -> alleles[0] ;
c2 = cupt -> alleles[1] ;
t1 = pedval(&c1) % 5 ;
t2 = pedval(&c2) % 5 ; // 0 and 5 are no good
if ((t1==0) && (t2 >0)) flip1(cupt, phasedmode, YES) ;
}
flipstrand(flipstrandname, snpmarkers, numsnps) ;
flipsnps(flipsnpname, snpmarkers, numsnps, phasedmode) ;
if (polarindex>=0) {
for (i=0; i<numsnps; i++) {
cupt = snpmarkers[i] ;
g = getgtypes(cupt, polarindex) ;
if (g==0) {
printf("polarizing %s\n", cupt -> ID) ;
flip1(cupt, NO, YES) ;
g = getgtypes(cupt, polarindex) ;
if (g!=2) fatalx("badbug\n") ;
}
if (g != 2) cupt -> ignore = YES ;
}
}
if (outputall) {
outfiles(snpoutfilename, indoutfilename, genooutfilename,
snpmarkers, indivmarkers, numsnps, numindivs, packout, ogmode) ;
printf("##end of convertf run (outputall mode)\n") ;
return 0 ;
}
if (poplistname != NULL)
{
ZALLOC(eglist, numindivs, char *) ;
numeg = loadlist(eglist, poplistname) ;
seteglist(indivmarkers, numindivs, poplistname);
for (i=0; i<numindivs; ++i) {
indx = indivmarkers[i] ;
if (indx -> affstatus == NO) indx -> ignore = YES ;
}
}
else
setstatus(indivmarkers, numindivs, "Case") ;
numsnps = rmsnps(snpmarkers, numsnps, deletesnpoutname) ;
numindivs = rmindivs(snpmarkers, numsnps, indivmarkers, numindivs) ;
if (killr2) {
nkill = killhir2(snpmarkers, numsnps, numindivs, r2physlim, r2genlim, r2thresh) ;
if (nkill>0) printf("killhir2. number of snps killed: %d\n", nkill) ;
}
if ( nhwfilter > 0 ) {
hwfilter(snpmarkers, numsnps, numindivs, nhwfilter, deletesnpoutname);
}
if ( xregionname ) {
excluderegions(xregionname, snpmarkers, numsnps, deletesnpoutname);
}
numvalidind = 0 ;
for (i=0; i<numindivs; ++i) {
indx = indivmarkers[i] ;
if (indx -> ignore) continue ;
if (numvalidgtind(snpmarkers, numsnps, i) ==0) {
indx -> ignore = YES ;
printf("no data for individual: %s\n", indx -> ID) ;
}
if (indx -> ignore == NO) ++numvalidind ;
}
if (maxmiss<0) maxmiss = (int) (maxmissfrac * (double) numvalidind+1) ;
printf("numvalidind: %5d maxmiss: %5d\n", numvalidind, maxmiss) ;
if (numvalidind == 0) fatalx("no valid samples!\n") ;
for (k=0; k<numsnps; ++k) {
if (maxmiss>numvalidind) break ;
cupt = snpmarkers[k] ;
t = numvalidind - numvalidgtypes(cupt) ;
// printf("zz %20s %4d %4d\n", cupt -> ID, t, numvalidind-t) ;
if (maxmiss < t) {
cupt -> ignore = YES ;
}
/**
if (numvalidind == t) {
printf("no data for snp: %s\n", cupt -> ID) ;
cupt -> ignore = YES ;
}
*/
}
if (fastdup) {
printf("fastdup set %d\n", fastdupnum) ;
if (fastdupnum > 0) {
setfastdupnum(fastdupnum) ;
setfastdupthresh(fastdupthresh, fastdupkill) ;
fastdupcheck(snpmarkers, indivmarkers, numsnps, numindivs) ;
}
}
if (decim>0) {
snpdecimate(snpmarkers, numsnps, decim, dmindis, dmaxdis) ;
}
outfiles(snpoutfilename, indoutfilename, genooutfilename,
snpmarkers, indivmarkers, numsnps, numindivs, packout, ogmode) ;
printf("##end of convertf run\n") ;
return 0 ;
}
int main(int argc, char **argv)
{
char **eglist ;
int numeg ;
int i, j, k, pos;
int *vv ;
SNP *cupt, *cupt2 ;
Indiv *indx ;
double y1, y2, y ;
int n0, n1, nkill ;
int nindiv = 0 ;
int nignore, numrisks = 1 ;
SNP **xsnplist ;
Indiv **xindlist ;
int *xindex ;
int nrows, ncols, m ;
double *XTX, *cc, *evecs, *ww ;
double *lambda ;
double *tvecs ;
int weightmode = NO ;
int t ;
double *xmean, *xfancy ;
double *ldmat = NULL, *ldmat2 = NULL;
double *ldvv = NULL, *ldvv2 = NULL, *vv2 = NULL ;
int chrom, numclear ;
double gdis ;
int outliter, numoutiter, *badlist, nbad ;
int a, b, n ;
FILE *outlfile ;
int xblock, blocksize=10000 ;
double *tblock ;
OUTLINFO *outpt ;
int *idperm, *vecind ; // for sort
readcommands(argc, argv) ;
printf("## smartrel version: %s\n", WVERSION) ;
packmode = YES ;
setomode(&outputmode, omode) ;
if (parname == NULL) return 0 ;
if (xchrom == (numchrom+1)) noxdata = NO ;
if (fstonly) {
printf("fstonly\n") ;
numeigs = 0 ;
numoutliter = 0 ;
numoutiter = 0 ;
outputname = NULL ;
snpeigname = NULL ;
}
if (fancynorm) printf("norm used\n\n") ;
else printf("no norm used\n\n") ;
nostatslim = MAX(nostatslim, 3) ;
outlfile = ofile = stdout;
if (outputname != NULL) openit(outputname, &ofile, "w") ;
if (outliername != NULL) openit(outliername, &outlfile, "w") ;
if (fstdetailsname != NULL) openit(fstdetailsname, &fstdetails, "w") ;
numsnps =
getsnps(snpname, &snpmarkers, 0.0, badsnpname, &nignore, numrisks) ;
numindivs = getindivs(indivname, &indivmarkers) ;
k = getgenos(genotypename, snpmarkers, indivmarkers,
numsnps, numindivs, nignore) ;
if (poplistname != NULL)
{
ZALLOC(eglist, numindivs, char *) ;
numeg = loadlist(eglist, poplistname) ;
seteglist(indivmarkers, numindivs, poplistname);
}
else
{
setstatus(indivmarkers, numindivs, NULL) ;
ZALLOC(eglist, MAXPOPS, char *) ;
numeg = makeeglist(eglist, MAXPOPS, indivmarkers, numindivs) ;
}
for (i=0; i<numeg; i++)
{
/* printf("%3d %s\n",i, eglist[i]) ; */
}
nindiv=0 ;
for (i=0; i<numindivs; i++)
{
indx = indivmarkers[i] ;
if(indx -> affstatus == YES) ++nindiv ;
}
for (i=0; i<numsnps; i++)
{
cupt = snpmarkers[i] ;
chrom = cupt -> chrom ;
if ((noxdata) && (chrom == (numchrom+1))) cupt-> ignore = YES ;
if (chrom == 0) cupt -> ignore = YES ;
if (chrom > (numchrom+1)) cupt -> ignore = YES ;
}
for (i=0; i<numsnps; i++)
{
cupt = snpmarkers[i] ;
pos = nnint(cupt -> physpos) ;
if ((xchrom>0) && (cupt -> chrom != xchrom)) cupt -> ignore = YES ;
if ((xchrom > 0) && (pos < lopos)) cupt -> ignore = YES ;
if ((xchrom > 0) && (pos > hipos)) cupt -> ignore = YES ;
if (cupt -> ignore) continue ;
if (numvalidgtx(indivmarkers, cupt, YES) <= 1)
{
printf("nodata: %20s\n", cupt -> ID) ;
cupt -> ignore = YES ;
}
}
if (killr2) {
nkill = killhir2(snpmarkers, numsnps, numindivs, r2physlim, r2genlim, r2thresh) ;
if (nkill>0) printf("killhir2. number of snps killed: %d\n", nkill) ;
}
ZALLOC(vv, numindivs, int) ;
numvalidgtallind(vv, snpmarkers, numsnps, numindivs) ;
for (i=0; i<numindivs; ++i) {
if (vv[i] == 0) {
indx = indivmarkers[i] ;
indx -> ignore = YES ;
}
}
free(vv) ;
numsnps = rmsnps(snpmarkers, numsnps, NULL) ; // rid ignorable snps
if (missingmode)
{
setmiss(snpmarkers, numsnps) ;
fancynorm = NO ;
}
if (weightname != NULL)
{
weightmode = YES ;
getweights(weightname, snpmarkers, numsnps) ;
}
if (ldregress>0)
{
ZALLOC(ldvv, ldregress*numindivs, double) ;
ZALLOC(ldvv2, ldregress*numindivs, double) ;
ZALLOC(vv2, numindivs, double) ;
ZALLOC(ldmat, ldregress*ldregress, double) ;
ZALLOC(ldmat2, ldregress*ldregress, double) ;
setidmat(ldmat, ldregress) ;
vst(ldmat, ldmat, 1.0e-6, ldregress*ldregress) ;
}
ZALLOC(xindex, numindivs, int) ;
ZALLOC(xindlist, numindivs, Indiv *) ;
ZALLOC(xsnplist, numsnps, SNP *) ;
if (popsizelimit > 0)
{
setplimit(indivmarkers, numindivs, eglist, numeg, popsizelimit) ;
}
nrows = loadindx(xindlist, xindex, indivmarkers, numindivs) ;
ncols = loadsnpx(xsnplist, snpmarkers, numsnps, indivmarkers) ;
printf("number of samples used: %d number of snps used: %d\n", nrows, ncols) ;
/**
cupt = xsnplist[0] ;
for (j=0; j<nrows; ++j) {
k = xindex[j] ;
g = getgtypes(cupt, k) ;
indx = indivmarkers[k] ;
t = indxindex(eglist, numeg, indx -> egroup) ;
printf("yy1 %20s %20s %20s %d %d %d\n", cupt ->ID, indx -> ID, indx -> egroup, j, k, g) ;
}
printf("yya: ") ; printimat(xindex, 1, nrows) ;
printf("zzindxa: %s\n", indivmarkers[230] -> egroup) ;
*/
/* printf("## nrows: %d ncols %d\n", nrows, ncols) ; */
ZALLOC(xmean, ncols, double) ;
ZALLOC(xfancy, ncols, double) ;
ZALLOC(XTX, nrows*nrows, double) ;
ZALLOC(evecs, nrows*nrows, double) ;
ZALLOC(tvecs, nrows*nrows, double) ;
ZALLOC(lambda, nrows, double) ;
ZALLOC(cc, nrows, double) ;
ZALLOC(ww, nrows, double) ;
ZALLOC(badlist, nrows, int) ;
blocksize = MIN(blocksize, ncols) ;
ZALLOC(tblock, nrows*blocksize, double) ;
// xfancy is multiplier for column xmean is mean to take off
// badlist is list of rows to delete (outlier removal)
numoutiter = 1 ;
if (numoutliter>=1)
{
numoutiter = numoutliter+1 ;
ZALLOC(outinfo, nrows, OUTLINFO *) ;
for (k=0; k<nrows; k++)
{
ZALLOC(outinfo[k], 1, OUTLINFO) ;
}
/* fprintf(outlfile, "##%18s %4s %6s %9s\n", "ID", "iter","eigvec", "score") ; */
}
for (outliter = 1; outliter <= numoutiter ; ++outliter) {
if (fstonly) {
setidmat(XTX, nrows) ;
vclear(lambda, 1.0, nrows) ;
break ;
}
if (outliter>1) {
ncols = loadsnpx(xsnplist, snpmarkers, numsnps, indivmarkers) ;
}
vzero(XTX, nrows*nrows) ;
vzero(tblock, nrows*blocksize) ;
xblock = 0 ;
vzero(xmean, ncols) ;
vclear(xfancy, 1.0, ncols) ;
for (i=0; i<ncols; i++)
{
cupt = xsnplist[i] ;
chrom = cupt -> chrom ;
getcolxz(cc, cupt, xindex, nrows, i, xmean, xfancy, &n0, &n1) ;
t = MIN(n0, n1) ;
if (t <= minallelecnt) {
cupt -> ignore = YES ;
vzero(cc, nrows) ;
}
if (weightmode)
{
vst(cc, cc, xsnplist[i] -> weight, nrows) ;
}
if (ldregress>0)
{
numclear = 0 ;
for (k=1; k<= ldregress; ++k)
{
j = i-k ;
if (j<0)
{
numclear = ldregress-k+1 ;
break ;
}
cupt2 = xsnplist[j] ;
if (cupt2 -> chrom != chrom) gdis = ldlimit + 1.0 ;
else gdis = cupt -> genpos - cupt2 -> genpos ;
if (gdis>=ldlimit)
{
numclear = ldregress-k+1 ;
break ;
}
}
if (numclear>0) clearld(ldmat, ldvv, ldregress, nrows, numclear) ;
ldreg(ldmat, ldmat2, cc, vv2, ldvv, ldvv2, ldregress, nrows) ;
copyarr(ldmat2, ldmat, ldregress*ldregress) ;
copyarr(vv2, cc, nrows) ;
copyarr(ldvv2, ldvv, ldregress*nrows) ;
}
copyarr(cc, tblock+xblock*nrows, nrows) ;
++xblock ;
/** this is the key code to parallelize */
if (xblock==blocksize)
{
domult(tvecs, tblock, xblock, nrows) ;
vvp(XTX, XTX, tvecs, nrows*nrows) ;
xblock = 0 ;
vzero(tblock, nrows*blocksize) ;
}
}
if (xblock>0)
{
domult(tvecs, tblock, xblock, nrows) ;
vvp(XTX, XTX, tvecs, nrows*nrows) ;
}
symit(XTX, nrows) ;
/**
a = 0; b=0 ;
printf("zz1 %12.6f ", XTX[a*nrows+b]) ;
a = nrows-1; b=nrows-1 ;
printf(" %12.6f %15.9g\n", XTX[a*nrows+b], asum(XTX, nrows*nrows)) ;
*/
if (verbose)
{
printdiag(XTX, nrows) ;
}
y = trace(XTX, nrows) / (double) (nrows-1) ;
if (isnan(y)) fatalx("bad XTX matrix\n") ;
/* printf("trace: %9.3f\n", y) ; */
if (y<=0.0) fatalx("XTX has zero trace (perhaps no data)\n") ;
vst(XTX, XTX, 1.0/y, nrows * nrows) ;
/// mean eigenvalue is 1
eigvecs(XTX, lambda, evecs, nrows) ;
// eigenvalues are in decreasing order
if (outliter > numoutliter) break ;
// last pass skips outliers
numoutleigs = MIN(numoutleigs, nrows-1) ;
nbad = ridoutlier(evecs, nrows, numoutleigs, outlthresh, badlist, outinfo) ;
if (nbad == 0) break ;
for (i=0; i<nbad; i++)
{
j = badlist[i] ;
indx = xindlist[j] ;
outpt = outinfo[j] ;
fprintf(outlfile, "REMOVED outlier %s iter %d evec %d sigmage %.3f\n", indx -> ID, outliter, outpt -> vecno, outpt -> score) ;
indx -> ignore = YES ;
}
nrows = loadindx(xindlist, xindex, indivmarkers, numindivs) ;
printf("number of samples after outlier removal: %d\n", nrows) ;
}
if (outliername != NULL) fclose(outlfile) ;
m = numgtz(lambda, nrows) ;
/* printf("matrix rank: %d\n", m) ; */
if (m==0) fatalx("no data\n") ;
/** smartrel code */
for (i=0; i<numeigs; i++) {
y = sqrt(lambda[i]) ;
vst(ww, evecs+i*nrows, y, nrows) ;
subouter(XTX, ww, nrows) ;
}
free(tvecs) ;
n = 0 ;
ZALLOC(vecind, nrows*nrows/2, int) ;
for (i=0; i<nrows; i++) {
for (j=i+1; j<nrows; j++) {
k = i*nrows + j ;
y1 = XTX[i*nrows+i] ;
y2 = XTX[j*nrows+j] ;
y = XTX[k]/sqrt(y1*y2) ;
y += 1/(double)(nrows-1);
if (y<relthresh) continue ;
vecind[n] = k ;
evecs[n] = -y ;
++n ;
}
}
free(XTX) ;
if (n==0) {
printf("## nothing above relthresh!\n") ;
printf("##end of smartrel run\n") ;
return 0 ;
}
ZALLOC(idperm, n, int) ;
sortit(evecs, idperm, n) ;
for (i=0; i<n; i++) {
j = idperm[i] ;
k = vecind[j] ;
a = k/nrows ;
b = k%nrows ;
printf("rel: %20s ", xindlist[a] ->ID) ;
printf("%20s ", xindlist[b] ->ID) ;
printf(" %9.3f", -evecs[i]) ;
printnl() ;
}
printf("##end of smartrel run\n") ;
return 0 ;
}
int main(int argc, char **argv)
{
int **snppos ;
int *snpindx ;
char **snpnamelist, **indnamelist ;
char **eglist ;
int lsnplist, lindlist, numeg ;
int i,j;
SNP *cupt, *cupt1, *cupt2, *cupt3 ;
Indiv *indx ;
double gpos1,gpos2,cpos1,cpos2,gd, cd, gd100 ;
double rthresh, zt ;
int mpflag, ret, nvalid;
int ch1, ch2 ;
int fmnum , lmnum ;
int num, n1, n2 ;
int nkill = 0 ;
int t, k ;
int nindiv = 0, e, f, lag=1 ;
double xc[9], xd[4], xc2[9] ;
double ychi, zscore, zthresh = 20.0 ;
double y1, y2 ;
int nignore, numrisks = 1 ;
char **genolist ;
int numgenolist ;
int maxmiss ;
malexhet = YES ; // convertf default is don't change the data
tersem = YES ; // no snp counts
readcommands(argc, argv) ;
setomode(&outputmode, omode) ;
packmode = YES ;
settersemode(tersem) ;
numsnps =
getsnps(snpname, &snpmarkers, 0.0, badsnpname, &nignore, numrisks) ;
for (i=0; i<numsnps; i++) {
if (xchrom == -1) break ;
cupt = snpmarkers[i] ;
if (cupt -> chrom != xchrom) cupt -> ignore = YES ;
if (cupt -> ignore) continue ;
t = nnint(cupt -> physpos) ;
if ( (t< lopos) || (t >hipos)) cupt -> ignore = YES ;
}
nignore = 0 ;
for (i=0; i<numsnps; i++) {
cupt = snpmarkers[i] ;
if (cupt -> chrom > maxchrom) cupt -> ignore = YES ;
if (cupt -> chrom < minchrom) cupt -> ignore = YES ;
if (cupt -> ignore) ++nignore ;
}
if (numsnps == nignore) fatalx("no valid snps\n") ;
/**
cupt = snpmarkers[0] ;
printf("zz2: %d %d %d %20s: %d\n", numsnps, nignore, cupt -> chrom, cupt -> ID, cupt -> ignore) ;
*/
numindivs = getindivs(indivname, &indivmarkers) ;
if (genotypelist!= NULL) {
getgenos_list(genotypelist, snpmarkers, indivmarkers,
numsnps, numindivs, nignore) ;
}
else {
setgenotypename(&genotypename, indivname) ;
getgenos(genotypename, snpmarkers, indivmarkers,
numsnps, numindivs, nignore) ;
}
if (outputall) {
outfiles(snpoutfilename, indoutfilename, genooutfilename,
snpmarkers, indivmarkers, numsnps, numindivs, packout, ogmode) ;
printf("##end of convertf run (outputall mode)\n") ;
return 0 ;
}
if (killr2) {
nkill = killhir2(snpmarkers, numsnps, numindivs, r2physlim, r2genlim, r2thresh) ;
if (nkill>0) printf("killhir2. number of snps killed: %d\n", nkill) ;
}
setstatus(indivmarkers, numindivs, "Case") ;
/******************************************************************/
/* removesubthreshold(indivmarkers, snpmarkers, numindiv, numsnps,
maxmissfracind, maxmissfracsnp); */
/******************************************************************/
if (fastdup) {
if (fastdupnum > 0) setfastdupnum(fastdupnum) ;
fastdupcheck(snpmarkers, indivmarkers, numsnps, numindivs) ;
}
if (decim>0) {
snpdecimate(snpmarkers, numsnps, decim, dmindis, dmaxdis) ;
}
outfiles(snpoutfilename, indoutfilename, genooutfilename,
snpmarkers, indivmarkers, numsnps, numindivs, packout, ogmode) ;
printf("##end of convertf run\n") ;
return 0 ;
}
