/* Function: p7_emit_FancyConsensus()
* Synopsis: Emit a fancier consensus with upper/lower case and N/X's.
* Incept: SRE, Fri May 14 09:33:10 2010 [Janelia]
*
* Purpose: Generate a consensus sequence for model <hmm>, consisting
* of the maximum probability residue in each match state;
* store this sequence in text-mode <sq> provided by the caller.
*
* If the probability of the consensus residue is less than
* <min_lower>, show an ``any'' residue (N or X) instead.
* If the probability of the consensus residue is $\geq$
* <min_lower> and less than <min_upper>, show the residue
* as lower case; if it is $\geq$ <min_upper>, show it as
* upper case.
*
* Returns: <eslOK> on success.
*
* Throws: <eslEINVAL> if the <sq> isn't in text mode.
*
* Xref: SRE:J6/59.
*/
int
p7_emit_FancyConsensus(const P7_HMM *hmm, float min_lower, float min_upper, ESL_SQ *sq)
{
int k, x;
float p;
char c;
int status;
if (! esl_sq_IsText(sq)) ESL_EXCEPTION(eslEINVAL, "p7_emit_FancyConsensus() expects a text-mode <sq>");
if ((status = esl_sq_GrowTo(sq, hmm->M)) != eslOK) return status;
for (k = 1; k <= hmm->M; k++)
{
if (hmm->mm && hmm->mm[k] == 'm') { //masked position, spit out the degenerate code
if ((status = esl_sq_CAddResidue(sq, tolower(esl_abc_CGetUnknown(hmm->abc))) ) != eslOK) return status;
} else {
p = esl_vec_FMax( hmm->mat[k], hmm->abc->K);
x = esl_vec_FArgMax(hmm->mat[k], hmm->abc->K);
if (p < min_lower) c = tolower(esl_abc_CGetUnknown(hmm->abc));
else if (p >= min_upper) c = toupper(hmm->abc->sym[x]);
else c = tolower(hmm->abc->sym[x]);
if ((status = esl_sq_CAddResidue(sq, c)) != eslOK) return status;
}
}
if ((status = esl_sq_CAddResidue(sq, '\0')) != eslOK) return status;
return eslOK;
}
/* Function: p7_oprofile_GetSSVEmissionScoreArray()
* Synopsis: Retrieve MSV residue emission scores from a
* profile into an array
*
* Purpose: Extract an implicitly 2D array of 8-bit int MSV residue
* emission scores from a profile <om>. <arr> must
* be allocated by the calling function to be of size
* ( om->abc->Kp * ( om->M + 1 )), and indexing into the array
* is done as [om->abc->Kp * i + c ] for character c at
* position i.
*
* In the dummy implementation, we need to convert from the
* float emission probabilities to 8-bit int scores. Conversion
* is based on code from the function mf_conversion in impl_sse's
* p7_oprofile.c
*
* Args: <om> - profile, containing emission information
* <arr> - preallocated array into which scores will be placed
*
* Returns: <eslOK> on success.
*
* Throws: (no abnormal error conditions)
*/
int
p7_oprofile_GetSSVEmissionScoreArray(const P7_OPROFILE *om, uint8_t *arr )
{
int M = om->M; /* length of the query */
int i, j;
float x;
float max = 0.0;
float scale;
uint8_t bias;
/* scale and bias required for float->8bit conversion */
scale = 3.0 / eslCONST_LOG2; /* scores in units of third-bits */
for (i = 0; i < om->abc->K; i++) max = ESL_MAX(max, esl_vec_FMax(om->rsc[i], (M+1)*2));
max = -1.0f * roundf(scale * -1.0 * max); //based on unbiased_byteify
bias = (max > 255.) ? 255 : (uint8_t) max;
for (i = 1; i <= om->M; i++) {
for (j=0; j<om->abc->Kp; j++) {
//based on p7_oprofile's biased_byteify()
x = -1.0f * roundf(scale * om->rsc[j][(i) * p7P_NR + p7P_MSC]);
arr[i*om->abc->Kp + j] = (x > 255. - bias) ? 255 : (uint8_t) (x + bias);
}
}
return eslOK;
}
float
esl_vec_FLogSum(float *vec, int n)
{
int x;
float max, sum;
max = esl_vec_FMax(vec, n);
sum = 0.0;
for (x = 0; x < n; x++)
if (vec[x] > max - 50.)
sum += expf(vec[x] - max);
sum = logf(sum) + max;
return sum;
}
