/** * int main(int argc, char **argv) * Main driver */ int main(int argc, char **argv) { ESL_GETOPTS *go = NULL; /* command line processing */ ESL_ALPHABET *abc = NULL; char *hmmfile = NULL; char *outhmmfile = NULL; P7_HMMFILE *hfp = NULL; FILE *outhmmfp; /* HMM output file handle */ P7_HMM *hmm = NULL; P7_BG *bg = NULL; int nhmm; double x; float KL; int status; char errbuf[eslERRBUFSIZE]; float average_internal_transitions[ p7H_NTRANSITIONS ]; int k; char errmsg[eslERRBUFSIZE]; /* Process the command line options. */ go = esl_getopts_Create(options); if (esl_opt_ProcessCmdline(go, argc, argv) != eslOK || esl_opt_VerifyConfig(go) != eslOK) { printf("Failed to parse command line: %s\n", go->errbuf); esl_usage(stdout, argv[0], usage); printf("\nTo see more help on available options, do %s -h\n\n", argv[0]); exit(1); } if (esl_opt_GetBoolean(go, "-h") == TRUE) { profillic_p7_banner(stdout, argv[0], banner); esl_usage(stdout, argv[0], usage); puts("\nOptions:"); esl_opt_DisplayHelp(stdout, go, 0, 2, 80); /* 0=docgroup, 2 = indentation; 80=textwidth*/ exit(0); } if (esl_opt_ArgNumber(go) != 2) { puts("Incorrect number of command line arguments."); esl_usage(stdout, argv[0], usage); printf("\nTo see more help on available options, do %s -h\n\n", argv[0]); exit(1); } if ((hmmfile = esl_opt_GetArg(go, 1)) == NULL) { puts("Failed to read <input hmmfile> argument from command line."); esl_usage(stdout, argv[0], usage); printf("\nTo see more help on available options, do %s -h\n\n", argv[0]); exit(1); } if ((outhmmfile = esl_opt_GetArg(go, 2)) == NULL) { puts("Failed to read <output hmmfile> argument from command line."); esl_usage(stdout, argv[0], usage); printf("\nTo see more help on available options, do %s -h\n\n", argv[0]); exit(1); } profillic_p7_banner(stdout, argv[0], banner); /* Initializations: open the input HMM file for reading */ status = p7_hmmfile_OpenE(hmmfile, NULL, &hfp, errbuf); if (status == eslENOTFOUND) p7_Fail("File existence/permissions problem in trying to open HMM file %s.\n%s\n", hmmfile, errbuf); else if (status == eslEFORMAT) p7_Fail("File format problem in trying to open HMM file %s.\n%s\n", hmmfile, errbuf); else if (status != eslOK) p7_Fail("Unexpected error %d in opening HMM file %s.\n%s\n", status, hmmfile, errbuf); /* Initializations: open the output HMM file for writing */ if ((outhmmfp = fopen(outhmmfile, "w")) == NULL) ESL_FAIL(status, errmsg, "Failed to open HMM file %s for writing", outhmmfile); /* Main body: read HMMs one at a time, print one line of stats */ printf("#\n"); printf("# %-4s %-20s %-12s %8s %8s %6s %6s %6s %6s %6s\n", "idx", "name", "accession", "nseq", "eff_nseq", "M", "relent", "info", "p relE", "compKL"); printf("# %-4s %-20s %-12s %8s %8s %6s %6s %6s %6s %6s\n", "----", "--------------------", "------------", "--------", "--------", "------", "------", "------", "------", "------"); nhmm = 0; while ((status = p7_hmmfile_Read(hfp, &abc, &hmm)) != eslEOF) { if (status == eslEOD) esl_fatal("read failed, HMM file %s may be truncated?", hmmfile); else if (status == eslEFORMAT) esl_fatal("bad file format in HMM file %s", hmmfile); else if (status == eslEINCOMPAT) esl_fatal("HMM file %s contains different alphabets", hmmfile); else if (status != eslOK) esl_fatal("Unexpected error in reading HMMs from %s", hmmfile); nhmm++; if (bg == NULL) bg = p7_bg_Create(abc); esl_vec_FSet(average_internal_transitions, p7H_NTRANSITIONS, 0.); for( k = 1; k < hmm->M; k++ ) { esl_vec_FAdd(average_internal_transitions, hmm->t[k], p7H_NTRANSITIONS); } // Match transitions esl_vec_FNorm(average_internal_transitions, 3); // Insert transitions esl_vec_FNorm(average_internal_transitions + 3, 2); // Delete transitions esl_vec_FNorm(average_internal_transitions + 5, 2); // Ok now set them. for( k = 1; k < hmm->M; k++ ) { esl_vec_FCopy( average_internal_transitions, p7H_NTRANSITIONS, hmm->t[k] ); } if ((status = p7_hmm_Validate(hmm, errmsg, 0.0001)) != eslOK) return status; if ((status = p7_hmmfile_WriteASCII(outhmmfp, -1, hmm)) != eslOK) ESL_FAIL(status, errmsg, "HMM save failed"); p7_MeanPositionRelativeEntropy(hmm, bg, &x); p7_hmm_CompositionKLDist(hmm, bg, &KL, NULL); printf("%-6d %-20s %-12s %8d %8.2f %6d %6.2f %6.2f %6.2f %6.2f\n", nhmm, hmm->name, hmm->acc == NULL ? "-" : hmm->acc, hmm->nseq, hmm->eff_nseq, hmm->M, p7_MeanMatchRelativeEntropy(hmm, bg), p7_MeanMatchInfo(hmm, bg), x, KL); /* p7_MeanForwardScore(hmm, bg)); */ p7_hmm_Destroy(hmm); } p7_bg_Destroy(bg); esl_alphabet_Destroy(abc); p7_hmmfile_Close(hfp); if (outhmmfp != NULL) fclose(outhmmfp); esl_getopts_Destroy(go); exit(0); }
/* Function: p7_GNull2_ByExpectation() * Synopsis: Calculate null2 model from posterior probabilities. * Incept: SRE, Thu Feb 28 09:52:28 2008 [Janelia] * * Purpose: Calculate the "null2" model for the envelope encompassed * by a posterior probability calculation <pp> for model * <gm>. Return the null2 odds emission probabilities * $\frac{f'{x}}{f{x}}$ in <null2>, which caller * provides as space for at least <alphabet->Kp> residues. * * The expectation method is applied to envelopes in * simple, well resolved regions (regions containing just a * single envelope, where no stochastic traceback * clustering was required). * * Make sure that the posterior probability matrix <pp> has * been calculated by the caller for only the envelope; thus * its rows are numbered <1..Ld>, for envelope <ienv..jenv> * of length <Ld=jenv-ienv+1>. * * Args: gm - profile, in any mode, target length model set to <L> * pp - posterior prob matrix, for <gm> against domain envelope <dsq+i-1> (offset) * null2 - RETURN: null2 odds ratios per residue; <0..Kp-1>; caller allocated space * * Returns: <eslOK> on success; <null2> contains the null2 scores. The 0 * row of <pp> has been used as temp space, and happens to contain * the expected frequency that each M,I,N,C,J state is used in this * <pp> matrix to generate residues. * * Throws: (no abnormal error conditions) */ int p7_GNull2_ByExpectation(const P7_PROFILE *gm, P7_GMX *pp, float *null2) { int M = gm->M; int Ld = pp->L; float **dp = pp->dp; float *xmx = pp->xmx; float xfactor; int x; /* over symbols 0..K-1 */ int i; /* over offset envelope dsq positions 1..Ld */ int k; /* over model M states 1..M, I states 1..M-1 */ /* Calculate expected # of times that each emitting state was used * in generating the Ld residues in this domain. * The 0 row in <wrk> is used to hold these numbers. */ esl_vec_FCopy(pp->dp[1], (M+1)*p7G_NSCELLS, pp->dp[0]); esl_vec_FCopy(pp->xmx+p7G_NXCELLS, p7G_NXCELLS, pp->xmx); for (i = 2; i <= Ld; i++) { esl_vec_FAdd(pp->dp[0], pp->dp[i], (M+1)*p7G_NSCELLS); esl_vec_FAdd(pp->xmx, pp->xmx+i*p7G_NXCELLS, p7G_NXCELLS); } /* Convert those expected #'s to log frequencies; these we'll use as * the log posterior weights. */ esl_vec_FLog(pp->dp[0], (M+1)*p7G_NSCELLS); esl_vec_FLog(pp->xmx, p7G_NXCELLS); esl_vec_FIncrement(pp->dp[0], (M+1)*p7G_NSCELLS, -log((float)Ld)); esl_vec_FIncrement(pp->xmx, p7G_NXCELLS, -log((float)Ld)); /* Calculate null2's log odds emission probabilities, by taking * posterior weighted sum over all emission vectors used in paths * explaining the domain. * This is dog-slow; a point for future optimization. */ xfactor = XMX(0,p7G_N); xfactor = p7_FLogsum(xfactor, XMX(0,p7G_C)); xfactor = p7_FLogsum(xfactor, XMX(0,p7G_J)); esl_vec_FSet(null2, gm->abc->K, -eslINFINITY); for (x = 0; x < gm->abc->K; x++) { for (k = 1; k < M; k++) { null2[x] = p7_FLogsum(null2[x], MMX(0,k) + p7P_MSC(gm, k, x)); null2[x] = p7_FLogsum(null2[x], IMX(0,k) + p7P_ISC(gm, k, x)); } null2[x] = p7_FLogsum(null2[x], MMX(0,M) + p7P_MSC(gm, k, x)); null2[x] = p7_FLogsum(null2[x], xfactor); } esl_vec_FExp (null2, gm->abc->K); /* now null2[x] = \frac{f_d(x)}{f_0(x)} for all x in alphabet, * 0..K-1, where f_d(x) are the ad hoc "null2" residue frequencies * for this envelope. */ /* make valid scores for all degeneracies, by averaging the odds ratios. */ esl_abc_FAvgScVec(gm->abc, null2); /* does not set gap, nonres, missing */ null2[gm->abc->K] = 1.0; /* gap character */ null2[gm->abc->Kp-2] = 1.0; /* nonresidue "*" */ null2[gm->abc->Kp-1] = 1.0; /* missing data "~" */ return eslOK; }