/* Function: MsaPwIdent
*
* Purpose:
*
* Args: *msa
*
* Returns: pairwise sequence identity of msa file
*/
float MsaPwIdent(MSA *msa)
{
float **idmx; /* identity matrix */
float sum, avgid;
int i, j;
MakeIdentityMx(msa->aseq, msa->nseq, &idmx);
sum=0.0;
for (i = 0; i < msa->nseq; i++)
for (j = 0; j < i; j++)
sum += idmx[i][j];
avgid = sum / (float) (msa->nseq * (msa->nseq-1)/2.0);
FMX2Free(idmx);
return avgid;
}
/**
* @brief Stripped down version of squid's alistat
*
*
* @param[in] prMSeq
* The alignment to analyse
* @param[in] bSampling
* For many sequences: samples from pool
* @param[in] bReportAll
* Report identities for all sequence pairs
*
* Don't have to worry about sequence case because our version of PairwiseIdentity is case insensitive
*/
void
AliStat(mseq_t *prMSeq, bool bSampling, bool bReportAll) {
/*
* bSampling = squid's do_fast
* bReportAll = squid's allreport
*/
float **ppdIdentMx; /* identity matrix (squid: imx) */
const int iNumSample = 1000; /* sample size (squid: nsample) */
MSA *msa; /* squid's alignment structure */
int small, large;
int bestj, worstj;
float sum;
float worst_worst, worst_best, best_best;
float avgid;
int i, j;
int nres; /* number of residues */
if (bSampling && bReportAll) {
Log(&rLog, LOG_WARN,
"Cannot report all and sample at the same time. Skipping %s()", __FUNCTION__);
return;
}
if (FALSE == prMSeq->aligned) {
Log(&rLog, LOG_WARN,
"Sequences are not aligned. Skipping %s()", __FUNCTION__);
return;
}
/* silence gcc warnings about uninitialized variables
*/
worst_worst = worst_best = best_best = 0.0;
bestj = worstj = -1;
/** mseq to squid msa
*
* FIXME code overlap with WriteAlignment. Make it a function and take
* code there (contains more comments) as template
*
*/
msa = MSAAlloc(prMSeq->nseqs,
/* derive alignment length from first seq */
strlen(prMSeq->seq[0]));
for (i=0; i<prMSeq->nseqs; i++) {
int key; /* MSA struct internal index for sequence */
char *this_name = prMSeq->sqinfo[i].name; /* prMSeq sequence name */
char *this_seq = prMSeq->seq[i]; /* prMSeq sequence */
SQINFO *this_sqinfo = &prMSeq->sqinfo[i]; /* prMSeq sequence name */
key = GKIStoreKey(msa->index, this_name);
msa->sqname[key] = sre_strdup(this_name, strlen(this_name));
/* setting msa->sqlen[idx] and msa->aseq[idx] */
msa->sqlen[key] = sre_strcat(&(msa->aseq[key]), msa->sqlen[key],
this_seq, strlen(this_seq));
if (this_sqinfo->flags & SQINFO_DESC) {
MSASetSeqDescription(msa, key, this_sqinfo->desc);
}
msa->nseq++;
}
nres = 0;
small = large = -1;
for (i = 0; i < msa->nseq; i++) {
int rlen; /* raw sequence length */
rlen = DealignedLength(msa->aseq[i]);
nres += rlen;
if (small == -1 || rlen < small) small = rlen;
if (large == -1 || rlen > large) large = rlen;
}
if (bSampling) {
avgid = AlignmentIdentityBySampling(msa->aseq, msa->alen,
msa->nseq, iNumSample);
} else {
float best, worst;
/* this might be slow...could use openmp inside squid */
MakeIdentityMx(msa->aseq, msa->nseq, &ppdIdentMx);
if (bReportAll) {
printf(" %-15s %5s %7s %-15s %7s %-15s\n",
"NAME", "LEN", "HIGH ID", "(TO)", "LOW ID", "(TO)");
printf(" --------------- ----- ------- --------------- ------- ---------------\n");
}
sum = 0.0;
worst_best = 1.0;
best_best = 0.0;
worst_worst = 1.0;
for (i = 0; i < msa->nseq; i++) {
worst = 1.0;
best = 0.0;
for (j = 0; j < msa->nseq; j++) {
/* closest seq to this one = best */
if (i != j && ppdIdentMx[i][j] > best) {
best = ppdIdentMx[i][j];
bestj = j;
}
if (ppdIdentMx[i][j] < worst) {
worst = ppdIdentMx[i][j];
worstj = j;
}
}
if (bReportAll) {
printf("* %-15s %5d %7.1f %-15s %7.1f %-15s\n",
msa->sqname[i], DealignedLength(msa->aseq[i]),
best * 100., msa->sqname[bestj],
worst * 100., msa->sqname[worstj]);
}
if (best > best_best) best_best = best;
if (best < worst_best) worst_best = best;
if (worst < worst_worst) worst_worst = worst;
for (j = 0; j < i; j++)
sum += ppdIdentMx[i][j];
}
avgid = sum / (float) (msa->nseq * (msa->nseq-1)/2.0);
if (bReportAll)
puts("");
FMX2Free(ppdIdentMx);
} /* else bSampling */
/* Print output
*/
if (msa->name != NULL)
printf("Alignment name: %s\n", msa->name);
/*printf("Format: %s\n", SeqfileFormat2String(afp->format));*/
printf("Number of sequences: %d\n", msa->nseq);
printf("Total # residues: %d\n", nres);
printf("Smallest: %d\n", small);
printf("Largest: %d\n", large);
printf("Average length: %.1f\n", (float) nres / (float) msa->nseq);
printf("Alignment length: %d\n", msa->alen);
printf("Average identity: %.2f%%\n", 100.*avgid);
if (! bSampling) {
printf("Most related pair: %.2f%%\n", 100.*best_best);
printf("Most unrelated pair: %.2f%%\n", 100.*worst_worst);
printf("Most distant seq: %.2f%%\n", 100.*worst_best);
}
/*
char *cs;
cs = MajorityRuleConsensus(msa->aseq, msa->nseq, msa->alen);
printf cs;
*/
MSAFree(msa);
}
