X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=bam_mcns.c;h=b2edfe130bc9c32d5adbc9a8230edd122b8bdb73;hb=82c492c6e173dddc6876304b5bd14119ae9566fa;hp=b69ec0f0abaceb6aa0408cb61a9adc51b0626765;hpb=7795f54fa8aed1e0b1edff61890e9fe25fdb7fe9;p=samtools.git

diff --git a/bam_mcns.c b/bam_mcns.c
index b69ec0f..b2edfe1 100644
--- a/bam_mcns.c
+++ b/bam_mcns.c
@@ -1,30 +1,64 @@
 #include <math.h>
 #include <stdlib.h>
+#include <stdio.h>
 #include "bam_mcns.h"
 
 #define MC_MIN_QUAL 13
+#define MC_AVG_ERR 0.007
 #define MC_MAX_SUMQ 3000
 #define MC_MAX_SUMQP 1e-300
+#define MC_MAX_EM_ITER 16
+#define MC_EM_EPS 1e-4
 
 struct __mc_aux_t {
-	int n, N;
-	int ref, alt;
+	int n, M;
+	int ref, alt, alt2;
 	double *q2p, *pdg; // pdg -> P(D|g)
-	int *qsum, *bcnt, rcnt[4];
+	double *phi;
+	double *z, *zswap; // aux for afs
+	double *CMk; // \binom{M}{k}
+	double *afs, *afs1; // afs: accumulative AFS; afs1: site posterior distribution
+	int *qsum, *bcnt;
 };
 
+void mc_init_prior(mc_aux_t *ma, int type, double theta)
+{
+	int i;
+	if (type == MC_PTYPE_COND2) {
+		for (i = 0; i <= 2 * ma->n; ++i)
+			ma->phi[i] = 2. * (i + 1) / (2 * ma->n + 1) / (2 * ma->n + 2);
+	} else if (type == MC_PTYPE_FLAT) {
+		for (i = 0; i <= ma->M; ++i)
+			ma->phi[i] = 1. / (ma->M + 1);
+	} else {
+		double sum;
+		for (i = 0, sum = 0.; i < 2 * ma->n; ++i)
+			sum += (ma->phi[i] = theta / (2 * ma->n - i));
+		ma->phi[2 * ma->n] = 1. - sum;
+	}
+}
+
 mc_aux_t *mc_init(int n) // FIXME: assuming diploid
 {
 	mc_aux_t *ma;
 	int i;
 	ma = calloc(1, sizeof(mc_aux_t));
-	ma->n = n; ma->N = 2 * n;
+	ma->n = n; ma->M = 2 * n;
 	ma->q2p = calloc(MC_MAX_SUMQ + 1, sizeof(double));
 	ma->qsum = calloc(4 * ma->n, sizeof(int));
 	ma->bcnt = calloc(4 * ma->n, sizeof(int));
 	ma->pdg = calloc(3 * ma->n, sizeof(double));
+	ma->phi = calloc(2 * ma->n + 1, sizeof(double));
+	ma->z = calloc(2 * ma->n + 1, sizeof(double));
+	ma->zswap = calloc(2 * ma->n + 1, sizeof(double));
+	ma->afs = calloc(2 * ma->n + 1, sizeof(double));
+	ma->afs1 = calloc(2 * ma->n + 1, sizeof(double));
+	ma->CMk = calloc(ma->M + 1, sizeof(double));
 	for (i = 0; i <= MC_MAX_SUMQ; ++i)
 		ma->q2p[i] = pow(10., -i / 10.);
+	for (i = 0; i <= ma->M; ++i)
+		ma->CMk[i] = exp(lgamma(ma->M+1) - lgamma(ma->M-i+1) - lgamma(i+1));
+	mc_init_prior(ma, MC_PTYPE_FULL, 1e-3); // the simplest prior
 	return ma;
 }
 
@@ -33,21 +67,23 @@ void mc_destroy(mc_aux_t *ma)
 	if (ma) {
 		free(ma->qsum); free(ma->bcnt);
 		free(ma->q2p); free(ma->pdg);
+		free(ma->phi);
+		free(ma->z); free(ma->zswap);
+		free(ma->afs); free(ma->afs1);
+		free(ma->CMk);
 		free(ma);
 	}
 }
 
-static void sum_err(int *n, const bam_pileup1_t **plp, mc_aux_t *ma)
+static int sum_err(int *n, const bam_pileup1_t **plp, mc_aux_t *ma)
 {
-	int i, j;
+	int i, j, tot = 0;
 	memset(ma->qsum, 0, sizeof(int) * 4 * ma->n);
 	memset(ma->bcnt, 0, sizeof(int) * 4 * ma->n);
-	memset(ma->rcnt, 0, sizeof(int) * 4);
 	for (j = 0; j < ma->n; ++j) {
-		int tmp, *qsum = ma->qsum + j * 4;
+		int *qsum = ma->qsum + j * 4;
 		int *bcnt = ma->bcnt + j * 4;
-		double r = drand48(), rr;
-		for (i = tmp = 0; i < n[j]; ++i) {
+		for (i = 0; i < n[j]; ++i) {
 			const bam_pileup1_t *p = plp[j] + i;
 			int q, b;
 			if (p->is_del || (p->b->core.flag&BAM_FUNMAP)) continue;
@@ -58,16 +94,10 @@ static void sum_err(int *n, const bam_pileup1_t **plp, mc_aux_t *ma)
 			if (b > 3) continue; // N
 			qsum[b] += q;
 			++bcnt[b];
-			++tmp;
-		}
-		if (tmp) {
-			for (i = 0, rr = 0.; i < 4; ++i) {
-				rr += (double)bcnt[i] / tmp;
-				if (rr >= r) break;
-			}
-			if (i < 4) ++ma->rcnt[i];
+			++tot;
 		}
 	}
+	return tot;
 }
 
 static void set_allele(int ref, mc_aux_t *ma)
@@ -81,9 +111,13 @@ static void set_allele(int ref, mc_aux_t *ma)
 	for (i = 1; i < 4; ++i) // insertion sort
 		for (j = i; j > 0 && sum[j] < sum[j-1]; --j)
 			tmp = sum[j], sum[j] = sum[j-1], sum[j-1] = tmp;
-	ma->ref = sum[3]&3; ma->alt = sum[2]&3;
-	if (ref == ma->alt) tmp = ma->ref, ma->ref = ma->alt, ma->alt = tmp;
-	// note that ma->ref might not be ref in case of triallele
+	ma->ref = sum[3]&3; ma->alt = sum[2]&3; ma->alt2 = -1;
+	if (ma->ref != ref) { // the best base is not ref
+		if (ref >= 0 && ref <= 3) { // ref is not N
+			if (ma->alt == ref) tmp = ma->ref, ma->ref = ma->alt, ma->alt = tmp; // then switch alt and ref
+			else ma->alt2 = ma->alt, ma->alt = ma->ref, ma->ref = ref; // then set ref as ref
+		} else ma->alt2 = ma->alt, ma->alt = ma->ref, ma->ref = sum[0]&3; // then set the weakest as ref
+	}
 }
 
 static void cal_pdg(mc_aux_t *ma)
@@ -101,40 +135,148 @@ static void cal_pdg(mc_aux_t *ma)
 			pdg[i] = pi[i] > MC_MAX_SUMQ? MC_MAX_SUMQP : ma->q2p[pi[i]];
 	}
 }
-
-double mc_freq0(int ref, int *n, const bam_pileup1_t **plp, mc_aux_t *ma, int *_ref, int *alt)
+// this calculates the naive allele frequency and Nielsen's frequency
+static double mc_freq0(const mc_aux_t *ma, double *_f)
 {
-	int i, acnt[4], j;
-	double fr = -1., f0 = -1.;
-	sum_err(n, plp, ma);
-	set_allele(ref, ma);
-	cal_pdg(ma);
-	acnt[0] = acnt[1] = acnt[2] = acnt[3] = 0;
-	for (i = 0; i < ma->n; ++i)
-		for (j = 0; j < 4; ++j)
-			acnt[j] += ma->bcnt[i * 4 + j];
-	if (ma->rcnt[ma->ref] + ma->rcnt[ma->alt]) // never used...
-		fr = (double)ma->rcnt[ref] / (ma->rcnt[ma->ref] + ma->rcnt[ma->alt]);
-	if (acnt[ma->ref] + acnt[ma->alt])
-		f0 = (double)acnt[ref] / (acnt[ma->ref] + acnt[ma->alt]);
-	*_ref = ma->ref; *alt = ma->alt;
-	return ma->n == 1 || fr < 0.? f0 : fr;
+	int i, cnt;
+	double f, f_nielsen, w_sum;
+	*_f = -1.;
+	for (i = cnt = 0, f = f_nielsen = w_sum = 0.; i < ma->n; ++i) {
+		int *bcnt = ma->bcnt + i * 4;
+		int x = bcnt[ma->ref] + bcnt[ma->alt];
+		if (x) {
+			double w, p;
+			++cnt;
+			f += (double)bcnt[ma->ref] / x;
+			p = (bcnt[ma->ref] - MC_AVG_ERR * x) / (1. - 2. * MC_AVG_ERR) / x;
+			w = 2. * x / (1. + x);
+			w_sum += w;
+			f_nielsen += p * w;
+		}
+	}
+	if (cnt) {
+		f_nielsen /= w_sum;
+		if (f_nielsen < 0.) f_nielsen = 0.;
+		if (f_nielsen > 1.) f_nielsen = 1.;
+		*_f = f_nielsen;
+		return f / cnt;
+	} else return -1.;
 }
-
-double mc_freq_iter(double f0, mc_aux_t *ma)
+// f0 is the reference allele frequency
+static double mc_freq_iter(double f0, const mc_aux_t *ma)
 {
 	double f, f3[3];
-	int i, j;
+	int i;
 	f3[0] = (1.-f0)*(1.-f0); f3[1] = 2.*f0*(1.-f0); f3[2] = f0*f0;
 	for (i = 0, f = 0.; i < ma->n; ++i) {
-		double up, dn, *pdg;
+		double *pdg;
 		pdg = ma->pdg + i * 3;
-		for (j = 1, up = 0.; j < 3; ++j)
-			up += j * pdg[j] * f3[j];
-		for (j = 0, dn = 0.; j < 3; ++j)
-			dn += pdg[j] * f3[j];
-		f += up / dn;
+		f += (pdg[1] * f3[1] + 2. * pdg[2] * f3[2])
+			/ (pdg[0] * f3[0] + pdg[1] * f3[1] + pdg[2] * f3[2]);
 	}
 	f /= ma->n * 2.;
 	return f;
 }
+
+int mc_call_gt(const mc_aux_t *ma, double f0, int k)
+{
+	double sum, g[3];
+	double max, f3[3], *pdg = ma->pdg + k * 3;
+	int q, i, max_i;
+	f3[0] = (1.-f0)*(1.-f0); f3[1] = 2.*f0*(1.-f0); f3[2] = f0*f0;
+	for (i = 0, sum = 0.; i < 3; ++i)
+		sum += (g[i] = pdg[i] * f3[i]);
+	for (i = 0, max = -1., max_i = 0; i < 3; ++i) {
+		g[i] /= sum;
+		if (g[i] > max) max = g[i], max_i = i;
+	}
+	max = 1. - max;
+	if (max < 1e-308) max = 1e-308;
+	q = (int)(-3.434 * log(max) + .499);
+	if (q > 99) q = 99;
+	return q<<2|max_i;
+}
+
+static void mc_cal_z(mc_aux_t *ma)
+{
+	double *z[2], *tmp, *pdg;
+	int i, j;
+	z[0] = ma->z;
+	z[1] = ma->zswap;
+	pdg = ma->pdg;
+	z[0][0] = 1.; z[0][1] = z[0][2] = 0.;
+	for (j = 0; j < ma->n; ++j) {
+		int max = (j + 1) * 2;
+		double p[3];
+		pdg = ma->pdg + j * 3;
+		p[0] = pdg[0]; p[1] = 2. * pdg[1]; p[2] = pdg[2];
+		z[1][0] = p[0] * z[0][0];
+		z[1][1] = p[0] * z[0][1] + p[1] * z[0][0];
+		for (i = 2; i <= max; ++i)
+			z[1][i] = p[0] * z[0][i] + p[1] * z[0][i-1] + p[2] * z[0][i-2];
+		if (j < ma->n - 1) z[1][max+1] = z[1][max+2] = 0.;
+		tmp = z[0]; z[0] = z[1]; z[1] = tmp;
+	}
+	if (z[0] != ma->z) memcpy(ma->z, z[0], sizeof(double) * (2 * ma->n + 1));
+}
+
+static double mc_add_afs(mc_aux_t *ma)
+{
+	int k, l;
+	double sum = 0., avg = 0.;
+	memset(ma->afs1, 0, sizeof(double) * (ma->M + 1));
+	mc_cal_z(ma);
+	for (k = 0; k <= ma->M; ++k) {
+		for (l = 0, sum = 0.; l <= ma->M; ++l)
+			sum += ma->phi[l] * ma->z[l];
+		ma->afs1[k] = ma->phi[k] * ma->z[k] / sum;
+	}
+	for (k = 0; k <= ma->M; ++k)
+		if (isnan(ma->afs1[k]) || isinf(ma->afs1[k])) return -1.;
+	for (k = 0, sum = avg = 0.; k <= ma->M; ++k) {
+		ma->afs[k] += ma->afs1[k];
+		sum += ma->afs1[k];
+		avg += k * ma->afs1[k];
+	}
+//	for (k = 0; k <= ma->M; ++k) printf("^%lg:%lg:%lg ", ma->z[k], ma->phi[k], ma->afs1[k]);
+	return avg / ma->M;
+}
+
+int mc_cal(int ref, int *n, const bam_pileup1_t **plp, mc_aux_t *ma, mc_rst_t *rst, int level)
+{
+	int i, tot;
+	memset(rst, 0, sizeof(mc_rst_t));
+	rst->f_em = rst->f_exp = -1.; rst->ref = rst->alt = -1;
+	// precalculation
+	tot = sum_err(n, plp, ma);
+	if (tot == 0) return 0; // no good bases
+	set_allele(ref, ma);
+	cal_pdg(ma);
+	// set ref/major allele
+	rst->ref = ma->ref; rst->alt = ma->alt; rst->alt2 = ma->alt2;
+	// calculate naive and Nielsen's freq
+	rst->f_naive = mc_freq0(ma, &rst->f_nielsen);
+	{ // calculate f_em
+		double flast = rst->f_naive;
+		for (i = 0; i < MC_MAX_EM_ITER; ++i) {
+			rst->f_em = mc_freq_iter(flast, ma);
+			if (fabs(rst->f_em - flast) < MC_EM_EPS) break;
+			flast = rst->f_em;
+		}
+	}
+	if (level >= 2) {
+		rst->f_exp = mc_add_afs(ma);
+		rst->p_ref = ma->afs1[ma->M];
+	}
+	return tot;
+}
+
+void mc_dump_afs(mc_aux_t *ma)
+{
+	int k;
+	fprintf(stderr, "[afs]");
+	for (k = 0; k <= ma->M; ++k)
+		fprintf(stderr, " %d:%.3lf", k, ma->afs[ma->M - k]);
+	fprintf(stderr, "\n");
+	memset(ma->afs, 0, sizeof(double) * (ma->M + 1));
+}