static int
get_capture_config(const tchar *config_file, struct capture_config *config,
int add_flags, const tchar *fs_source_path)
{
int ret;
tchar *tmp_config_file = NULL;
memset(config, 0, sizeof(*config));
/* For WIMBoot capture, check for default capture configuration file
* unless one was explicitly specified. */
if (!config_file && (add_flags & WIMLIB_ADD_FLAG_WIMBOOT)) {
/* XXX: Handle loading file correctly when in NTFS volume. */
size_t len = tstrlen(fs_source_path) +
tstrlen(wimboot_cfgfile);
tmp_config_file = MALLOC((len + 1) * sizeof(tchar));
struct stat st;
tsprintf(tmp_config_file, T("%"TS"%"TS),
fs_source_path, wimboot_cfgfile);
if (!tstat(tmp_config_file, &st)) {
config_file = tmp_config_file;
add_flags &= ~WIMLIB_ADD_FLAG_WINCONFIG;
} else {
WARNING("\"%"TS"\" does not exist.\n"
" Using default capture configuration!",
tmp_config_file);
}
}
if (add_flags & WIMLIB_ADD_FLAG_WINCONFIG) {
/* Use Windows default. */
if (config_file)
return WIMLIB_ERR_INVALID_PARAM;
ret = read_capture_config(T("wincfg"), wincfg,
sizeof(wincfg) - 1, config);
} else if (config_file) {
/* Use the specified configuration file. */
ret = read_capture_config(config_file, NULL, 0, config);
} else {
/* ... Or don't use any configuration file at all. No files
* will be excluded from capture, all files will be compressed,
* etc. */
ret = 0;
}
FREE(tmp_config_file);
return ret;
}
// ----------------------------------------------------------------------------
// fefind
// Tries to find first free file name with specified extension (adding numbers
// to the name until 15).
// ----------------------------------------------------------------------------
TCHAR *fefind(TCHAR *fname, TCHAR *ext, TCHAR *newname)
{
TCHAR name[MAXS+1], newext[MAXS+1], *s;
struct _stat fst;
int ii;
xstrncpy(name, fname, MAXS);
if ((s = strrchr(name, T('.'))) != NULL) *s = T('\0');
xstrncpy(newname, name, MAXS);
xstrncat(newname, ext, MAXS);
for (ii = 0; ii <= 15; ii++) { // append 1..15 to output file name
if (tstat(newname, &fst) < 0) break; // file not found
snprintf(newext, MAXS, T("-%d%s"), ii+1, ext);
xstrncpy(newname, name, MAXS);
xstrncat(newname, newext, MAXS);
}
if (ii > 15) {
xstrncpy(newname, name, MAXS);
xstrncat(newname, ext, MAXS);
return NULL;
}
return newname;
} /* fefind */
float runOnePathway(char *line, pathwaySubject **sub, lrModel *model,paParams *pp, int writeFile)
{
char pathwayName[1000], pathwayURL[1000], gene[MAX_LOCI][50], scoreFileName[1000], outputFileName[1000], thisGene[50];
FILE *fs, *fo;
int nGene, missing[MAX_LOCI], s, g, n[2], i, cc, c,t,nValidGenes;
float sigma_x[2], sigma_x2[2], mean[2], var[2], SLP, SE, tval, s2, score;
double p, pathway_p;
fo = 0;
if (sscanf(line, "%s %s %[^\n]", pathwayName, pathwayURL, rest) != 3)
return 0;
for (nGene = 0; strcpy(line, rest), *rest = '\0', sscanf(line, "%s %[^\n]", gene[nGene], rest) >= 1; ++nGene)
;
for (s = 0; s < pp->nSub; ++s)
totScore[s] = 0;
nValidGenes = 0;
for (g = 0; g < nGene; ++g)
{
if (pp->scoreTableFile)
{
std::map<std::string, FILEOFFSET>::const_iterator geneIter = pp->geneIndex.find(gene[g]);
if (geneIter == pp->geneIndex.end())
missing[g] = 1;
else
{
fseek(pp->scoreTableFile, geneIter->second, SEEK_SET);
assert(fscanf(pp->scoreTableFile, "%s", thisGene) == 1);
if (strcmp(thisGene, gene[g]))
{
#if 0
dcerror(1, "Problem finding %s in %s. Tried to seek to position %" PRId64 " but got this: %s\n",
pp->scoreTableFileName, gene[g], geneIter->second, thisGene);
#else
dcerror(1, "Problem finding %s in %s. Tried to seek to position %uld but got this: %s\n",
pp->scoreTableFileName, gene[g], (unsigned long)geneIter->second, thisGene);
#endif
exit(1);
}
missing[g] = 0;
++nValidGenes;
for (s = 0; s < pp->nSub; ++s)
{
if (fscanf(pp->scoreTableFile, "%f", &sub[s]->score[g]) != 1)
{
dcerror(1, "Not enough entries in %s for gene %s\n", pp->scoreTableFileName, gene[g]); exit(1);
}
totScore[s] += sub[s]->score[g];
}
}
}
else
{
sprintf(scoreFileName, "%s%s%s", pp->scoreFilePrefix, gene[g], pp->scoreFileSuffix);
fs = fopen(scoreFileName, "r");
if (fs == 0)
missing[g] = 1;
else
{
missing[g] = 0;
++nValidGenes;
for (s = 0; s < pp->nSub; ++s)
{
if (!fgets(line, 1000, fs))
{
dcerror(1, "Not enough lines in scores file %s\n", scoreFileName); exit(1);
}
if (sscanf(line, "%s %d %f", sub[s]->id, &sub[s]->cc, &sub[s]->score[g]) != 3)
{
dcerror(1, "Not enough entries on this line in scores file %s:\n%s\n", scoreFileName, line); exit(1);
}
totScore[s] += sub[s]->score[g];
}
fclose(fs);
}
}
}
for (s = 0; s < pp->nSub; ++s)
totScore[s] /= nValidGenes; // use average score not total score
if (writeFile)
{
if (pp->summaryOutputFile != 0)
fprintf(pp->summaryOutputFile, "%s\t%s\t", pathwayName, pathwayURL);
sprintf(line, "%s%s%s", pp->outputFilePrefix, pathwayName, pp->outputFileSuffix);
fo = fopen(line, "w");
if (fo == 0)
{
dcerror(2, "Could not open output file %s\n", line);
}
else
fprintf(fo, "%s\n%s\n\n", pathwayName, pathwayURL);
}
if (pp->do_ttest)
{
for (i = 0; i < 2; ++i)
sigma_x[i] = sigma_x2[i] = n[i] = 0;
for (s = 0; s < pp->nSub; ++s)
{
cc = sub[s]->cc;
++n[cc];
score = totScore[s];
sigma_x[cc] += score;
sigma_x2[cc] += score * score;
}
for (i = 0; i < 2; ++i)
{
var[i] = (sigma_x2[i] - sigma_x[i] * sigma_x[i] / n[i]) / (n[i] - 1);
mean[i] = sigma_x[i] / n[i];
}
s2 = ((n[0] - 1)*var[0] + (n[1] - 1)*var[1]) / (n[0] + n[1] - 2);
SE = sqrt(s2*(1 / (float)n[0] + 1 / (float)n[1]));
if (SE == 0)
tval = 0;
else
tval = (mean[1] - mean[0]) / SE;
pathway_p = tstat(tval, n[0] + n[1] - 2.0) / 2; // one-tailed
SLP = log10(2 * pathway_p)*(mean[0] >= mean[1] ? 1 : -1);
if (fo != NULL)
{
fprintf(fo, " Controls Cases \n"
"N %9d %9d\n"
"Mean score %9.3f %9.3f\n"
"t (%d df) = %6.3f\n"
"p = %10.8f\n"
"SLP = %8.2f (signed log10(p), positive if cases score higher than controls)\n",
n[0], n[1], mean[0], mean[1], n[0] + n[1] - 2, tval, 2 * pathway_p, SLP);
if (pp->summaryOutputFile != 0)
fprintf(pp->summaryOutputFile, "%f\t", SLP);
if (SLP > pp->geneLevelOutputThreshold)
{
fprintf(fo, "\n\nSLPs for individual genes above threshold:\n");
for (g = 0; g < nGene; ++g)
{
if (missing[g] == 1)
continue;
for (i = 0; i < 2; ++i)
sigma_x[i] = sigma_x2[i] = n[i] = 0;
for (s = 0; s < pp->nSub; ++s)
{
cc = sub[s]->cc;
++n[cc];
score = sub[s]->score[g];
sigma_x[cc] += score;
sigma_x2[cc] += score * score;
}
for (i = 0; i < 2; ++i)
{
var[i] = (sigma_x2[i] - sigma_x[i] * sigma_x[i] / n[i]) / (n[i] - 1);
mean[i] = sigma_x[i] / n[i];
}
s2 = ((n[0] - 1)*var[0] + (n[1] - 1)*var[1]) / (n[0] + n[1] - 2);
SE = sqrt(s2*(1 / (float)n[0] + 1 / (float)n[1]));
if (SE == 0)
tval = 0;
else
tval = (mean[1] - mean[0]) / SE;
p = tstat(tval, n[0] + n[1] - 2.0) / 2; // one-tailed
SLP = log10(2 * p)*(mean[0] >= mean[1] ? 1 : -1);
if (SLP > pp->geneLevelOutputThreshold)
fprintf(fo, "%s %8.2f\n", gene[g], SLP);
}
fprintf(fo, "\n\nList of genes for which no score file was found:\n");
for (g = 0; g < nGene; ++g)
if (missing[g] == 1)
fprintf(fo, "%s\n", gene[g]);
}
}
if (mean[0] > mean[1])
pathway_p = 1.0 - pathway_p;
}
if(pp->do_lrtest||pp->numTestFiles>0)
fillModelWithVars(model,pp->nSub,pp,pp->scoreCol);
if (pp->do_lrtest)
{
SLP = do_onetailed_LRT(fo, model, pp);
if (writeFile &&pp->summaryOutputFile != 0)
fprintf(pp->summaryOutputFile, "%f\t", SLP);
}
if (pp->numTestFiles > 0)
{
for (t = 0; t < pp->numTestFiles; ++t)
{
SLP = runTestFile(fo, pp->testFiles[t].fn, model, pp);
if (writeFile &&pp->summaryOutputFile != 0)
fprintf(pp->summaryOutputFile, "%f\t", SLP);
}
}
if (fo)
fclose(fo);
if (writeFile && pp->summaryOutputFile != 0)
fprintf(pp->summaryOutputFile, "\n");
return SLP;
}
double do_score_onetailed_ttest(FILE *fo, float *score, subject **sub, int nsub, par_info *pi, sa_par_info *spi, float cc_freq[2][MAX_LOCI], float cc_count[2][MAX_LOCI], int max_cc[2], float *weight, float *missing, int *rarer)
{
int s, i, n[2], cc, pl, l, j;
float sigma_x[2], sigma_x2[2], mean[2], var[2], tval, SE, s2, rfreq, fscore, z, total_score;
double p, pz;
#ifndef USEMLP
float SLP;
#endif
for (i = 0; i<2; ++i)
sigma_x[i] = sigma_x2[i] = n[i] = 0;
for (s = 0; s<nsub; ++s)
{
cc = sub[s]->cc;
++n[cc];
sigma_x[cc] += score[s];
sigma_x2[cc] += score[s] * score[s];
}
for (i = 0; i<2; ++i)
{
if (spi->use_cc_freqs[i])
{
total_score = 0;
n[i] = max_cc[i];
var[i] = 0;
for (pl = 0; pl<pi->n_loci_to_use; ++pl)
{
float tempvar = 0;
float ngen[3]; // number of typed subjects who are AA,AB,BB
l = pi->loci_to_use[pl];
sigma_x2[i] = 0;
sigma_x[i] = 0;
if (rarer[l] == 2)
rfreq = cc_freq[i][l];
// assume supplied frequency is of alt allele, i.e. allele 2
// but score will be added to by rarer allele, even if this is allele 1
else
rfreq = 1 - cc_freq[i][l];
// as actual counts missing, assume HWE
ngen[0] = (1 - rfreq)*(1 - rfreq)*cc_count[i][l];
ngen[1] = 2 * rfreq*(1 - rfreq)*cc_count[i][l];
ngen[2] = rfreq*rfreq*cc_count[i][l];
fscore = weight[l]; // average score for each subject for this locus
sigma_x[i] += ngen[1] * fscore; // AB
sigma_x2[i] += ngen[1] * fscore*fscore;
sigma_x[i] += ngen[2] * 2 * fscore; // BB
sigma_x2[i] += ngen[2] * 4 * fscore*fscore;
fscore = missing[l]; // as if missing scores were also independent
sigma_x[i] += (max_cc[i] - cc_count[i][l])*fscore;
sigma_x2[i] += (max_cc[i] - cc_count[i][l])*fscore*fscore;
tempvar = (sigma_x2[i] - sigma_x[i] * sigma_x[i] / n[i]) / (n[i] - 1);
var[i] += tempvar; // add variances due to each marker
total_score += sigma_x[i];
}
mean[i] = total_score / n[i];
}
else
{
var[i] = (sigma_x2[i] - sigma_x[i] * sigma_x[i] / n[i]) / (n[i] - 1);
mean[i] = sigma_x[i] / n[i];
}
}
s2 = ((n[0] - 1)*var[0] + (n[1] - 1)*var[1]) / (n[0] + n[1] - 2);
SE = sqrt(s2*(1 / (float)n[0] + 1 / (float)n[1]));
if (SE == 0)
tval = 0;
else
tval = (mean[1] - mean[0]) / SE;
p = tstat(tval, n[0] + n[1] - 2.0) / 2; // one-tailed
SLP = log10(2 * p)*(mean[0] >= mean[1] ? 1 : -1);
if (fo != NULL)
fprintf(fo, " Controls Cases \n"
"N %9d %9d\n"
"Mean score %9.3f %9.3f\n"
"SD %9.3f %9.3f\n"
"t (%d df) = %6.3f\n"
"p = %10.8f\n"
"SLP = %8.2f (signed log10(p), positive if cases score higher than controls)\n",
n[0], n[1], mean[0], mean[1], sqrt(var[0]), sqrt(var[1]), n[0] + n[1] - 2, tval, 2 * p, SLP);
// I am writing SD because it will allow me to combine statistics later
if (mean[0]>mean[1])
p = 1.0 - p;
#ifdef ALLOWUNEQUALVARIANCES
SE = sqrt(var[0] / n[0] + var[1] / n[1]);
if (SE == 0)
z = 0;
else
z = (mean[1] - mean[0]) / SE;
pz = one_tailed_p_norm(z);
fprintf(fo, "Comparison of means not assuming equal variances, standardised normal deviate z = %5.3f\np = %10.8f (one-tailed)\n",
z, pz);
if (mean[0] >= mean[1])
SLP = log10(2 * (1 - pz));
else
SLP = -log10(2 * pz);
fprintf(fo, "SLP = %8.2f (signed log10(2p), positive if cases score higher than controls)\n",
SLP);
#endif
return p; // return one-tailed p
}
