static inline void gap_push(gap_stack_t *stack, int i, bwtint_t k, bwtint_t l, int n_mm, int n_gapo, int n_gape,
int state, int is_diff, const gap_opt_t *opt)
{
int score;
gap_entry_t *p;
gap_stack1_t *q;
score = aln_score(n_mm, n_gapo, n_gape, opt);
q = stack->stacks + score;
if (q->n_entries == q->m_entries) {
q->m_entries = q->m_entries? q->m_entries<<1 : 4;
q->stack = (gap_entry_t*)realloc(q->stack, sizeof(gap_entry_t) * q->m_entries);
}
p = q->stack + q->n_entries;
p->info = (u_int32_t)score<<21 | i;
p->k = k;
p->l = l;
p->n_mm = n_mm;
p->n_gapo = n_gapo;
p->n_gape = n_gape;
p->state = state;
p->last_diff_pos = is_diff? i : 0;
++(q->n_entries);
++(stack->n_entries);
if (stack->best > score) stack->best = score;
}
/* Calculate one sequence's contribution to the alignment score */
double marginal_score(glam2_scorer *s, glam2_aln *aln,
int seq, const fasta *f) {
double score = aln->score;
unalign(aln, seq, f);
score -= aln_score(s, aln);
realign(aln, seq, f);
return score;
}
void update_aln(glam2_aln *aln, data *d, const double temperature) {
assert(aln->seq_num > 0);
if (rand_dbl(d->a.column_sample_rate + 1) < 1) {
const int seq_pick = rand_int(aln->seq_num);
if (d->a.profile)
fprintf(d->out, "site sample: seq=%d\n", seq_pick);
site_sample(aln, seq_pick, d, temperature);
} else
column_sample(aln, d, temperature);
aln->score = aln_score(&d->scorer, aln);
}
/* get a random starting alignment */
void start_aln(glam2_aln *aln, data *d) {
int i;
#if 0
aln->width = d->a.min_width; /* ?? initial number of columns */
aln->width = sqrt(d->a.max_width * d->a.min_width); /* geometric mean */
#endif
aln->width = d->a.init_width;
aln_zero(aln);
SHUFFLE(d->seq_order, aln->seq_num);
for (i = 0; i < aln->seq_num; ++i)
site_sample(aln, d->seq_order[i], d, 1);
aln->score = aln_score(&d->scorer, aln);
}
gap_stack_t *gap_init_stack(int max_mm, int max_gapo, int max_gape, const gap_opt_t *opt)
{
int i;
gap_stack_t *stack;
stack = (gap_stack_t*)calloc(1, sizeof(gap_stack_t));
stack->n_stacks = aln_score(max_mm+1, max_gapo+1, max_gape+1, opt);
stack->stacks = (gap_stack1_t*)calloc(stack->n_stacks, sizeof(gap_stack1_t));
for (i = 0; i != stack->n_stacks; ++i) {
gap_stack1_t *p = stack->stacks + i;
p->m_entries = 4;
p->stack = (gap_entry_t*)calloc(p->m_entries, sizeof(gap_entry_t));
}
return stack;
}
static tmap_map1_aux_stack_t*
tmap_map1_aux_stack_reset(tmap_map1_aux_stack_t *stack,
int32_t max_mm, int32_t max_gapo, int32_t max_gape,
const tmap_map_opt_t *opt)
{
int32_t i;
//int32_t i, j;
int32_t n_bins_needed = 0;
// move to the beginning of the memory pool
stack->entry_pool_i = 0;
stack->best_score = INT32_MAX;
if(TMAP_MAP1_AUX_STACK_TOO_BIG < stack->entry_pool_length) {
tmap_map1_aux_stack_destroy_helper(stack, 0);
tmap_map1_aux_stack_init_helper(stack);
}
// clear the bins
for(i=0;i<stack->n_bins;i++) {
/*
for(j=0;j<stack->bins[i].n_entries;j++) {
stack->bins[i].entries[j] = NULL;
}
*/
stack->bins[i].n_entries = 0;
}
// resize the bins if necessary
n_bins_needed = aln_score(max_mm+1, max_gapo+1, max_gape+1, opt);
if(stack->n_bins < n_bins_needed) {
// realloc
tmap_roundup32(n_bins_needed);
stack->bins = tmap_realloc(stack->bins, sizeof(tmap_map1_aux_bin_t) * n_bins_needed, "stack->bins");
// initialize
for(i=stack->n_bins;i<n_bins_needed;i++) {
stack->bins[i].n_entries = stack->bins[i].m_entries = 0;
stack->bins[i].entries = NULL;
}
stack->n_bins = n_bins_needed;
}
stack->n_entries = 0;
return stack;
}
tmap_map_sams_t *
tmap_map1_aux_core(tmap_seq_t *seq, tmap_index_t *index,
tmap_bwt_match_hash_t *hash,
tmap_bwt_match_width_t *width, tmap_bwt_match_width_t *seed_width, tmap_map_opt_t *opt,
tmap_map1_aux_stack_t *stack, int32_t seed2_len)
{
int32_t max_mm = opt->max_mm, max_gapo = opt->max_gapo, max_gape = opt->max_gape, seed_max_diff = opt->seed_max_diff;
int32_t best_score, next_best_score;
int32_t best_cnt = 0;
int32_t i, j, num_n = 0;
int32_t max_edit_score;
tmap_bwt_match_occ_t match_sa_start;
tmap_string_t *bases=NULL;
tmap_map_sams_t *sams = NULL;
int32_t max_diff, best_diff;
tmap_bwt_int_t k, l;
tmap_refseq_t *refseq = index->refseq;
tmap_bwt_t *bwt = index->bwt;
tmap_sa_t *sa = index->sa;
tmap_map1_aux_occ_t *occs = NULL;
max_edit_score = opt->pen_mm;
//if(max_edit_score < opt->pen_gapo + opt->pen_gape) max_edit_score = opt->pen_gapo + opt->pen_gape;
//if(max_edit_score < opt->pen_gape) max_edit_score = opt->pen_gape;
bases = tmap_seq_get_bases(seq);
/*
fputc('\n', stderr);
for(i=0;i<bases->l;i++) {
fputc("ACGTN"[(int)bases->s[i]], stderr);
}
fputc('\n', stderr);
*/
// the maximum # of differences
if(bases->l <= TMAP_MAP_OPT_MAX_DIFF_READ_LENGTH) {
best_diff = max_diff = opt->max_diff_table[bases->l];
}
else {
best_diff = max_diff = opt->max_diff_table[TMAP_MAP_OPT_MAX_DIFF_READ_LENGTH];
}
// bound differenes by the maximum # of differences
if(max_diff < max_mm) max_mm = max_diff;
if(max_diff < max_gapo) max_gapo = max_diff;
//if(max_diff < max_gape) max_gape = max_diff;
best_score = next_best_score = aln_score(max_mm+1, max_gapo+1, max_gape+1, opt);
// check whether there are too many N
for(j=bases->l-seed2_len,num_n=0;j<bases->l;j++) {
if(3 < bases->s[j]) {
num_n++;
}
}
if(max_mm < num_n || max_diff < num_n) {
return tmap_map_sams_init(NULL);
}
// initialize
sams = tmap_map_sams_init(NULL);
occs = NULL;
match_sa_start.offset = 0;
match_sa_start.hi = 0;
match_sa_start.k = 0;
match_sa_start.l = bwt->seq_len;
stack = tmap_map1_aux_stack_reset(stack, max_mm, max_gapo, max_gape, opt); // reset stack
tmap_map1_aux_stack_push(stack, bases->l, &match_sa_start, 0, 0, 0, STATE_M, 0, NULL, opt);
while(0 < tmap_map1_aux_stack_size(stack) && tmap_map1_aux_stack_size(stack) < opt->max_entries) {
tmap_map1_aux_stack_entry_t *e = NULL;
int32_t len=-1;
int32_t n_seed_mm=0, offset, width_cur_i;
const uint8_t *str=NULL;
int32_t sam_found, m;
tmap_bwt_match_width_t *width_cur = NULL;
const tmap_bwt_match_width_t *seed_width_cur = NULL;
tmap_bwt_match_occ_t match_sa_cur, match_sa_next[4];
// get the best entry
e = tmap_map1_aux_stack_pop(stack);
// bound with best score
if(best_score + max_edit_score < e->score) {
break; // no need to continue
}
// some more information
match_sa_cur = e->match_sa;
// check if we have too many edits
m = max_diff - (e->n_mm + e->n_gapo + e->n_gape);
if(m < 0) {
continue; // too many edits
}
// get the rest of the information
offset = e->offset; // zero-based
str = (uint8_t*)bases->s;
len = bases->l;
width_cur = width;
width_cur_i = seed2_len - (len - offset);
if(NULL != seed_width) {
seed_width_cur = seed_width;
n_seed_mm = seed_max_diff - (e->n_mm + e->n_gapo + e->n_gape); // consider only mismatches in the seed
}
else {
seed_width_cur = NULL;
}
if(0 < width_cur_i && m < width_cur[width_cur_i-1].bid) { // too many edits
continue;
}
// check whether a sam is found
sam_found = 0;
if(len - seed2_len == offset) {
sam_found = 1;
}
else if(max_mm == e->n_mm // no mismatches from any state
&& ((e->state == STATE_M && max_gapo == e->n_gapo) // in STATE_M but no more gap opens
|| (e->state != STATE_M && max_gape == e->n_gape))) { // in STATE_I/STATE_D but no more extensions
if(0 < tmap_bwt_match_hash_exact_alt_reverse(bwt, offset, str, &match_sa_cur, hash)) { // the alignment must match exactly to sam
sam_found = 2;
}
else {
continue; // no sam, skip
}
}
if(0 < sam_found) { // alignment found
// check for duplicates
if(0 < sams->n) {
for(i=0;i<sams->n;i++) {
// check contained
if(match_sa_cur.k <= occs[i].k
&& occs[i].k <= match_sa_cur.l) { // MK <= SK <= ML
if(occs[i].l <= match_sa_cur.l) { // MK <= SK <= SL <= ML
// Want (SK - MK) + (ML - SL)
k = occs[i].k - match_sa_cur.k; // (SK - MK)
k += match_sa_cur.l - occs[i].l; // (ML - SL)
occs[i].l = match_sa_cur.l; // Make SL = ML
}
else { // MK <= SK <= ML <= SL
k = occs[i].k - match_sa_cur.k; // (SK - MK)
}
occs[i].k = match_sa_cur.k; // Make SK = MK
break;
}
else if(match_sa_cur.k <= occs[i].l
&& occs[i].l <= match_sa_cur.l) { // MK <= SL <= ML
if(match_sa_cur.k <= occs[i].k) { // MK <= SK <= SL <= ML
// Want (SK - MK) + (ML - SL)
k = occs[i].k - match_sa_cur.k; // (SK - MK)
k += match_sa_cur.l - occs[i].l; // (ML - SL)
occs[i].k = match_sa_cur.k; // Make SK = MK
}
else { // SK <= MK <= SL <= ML
k = match_sa_cur.l - occs[i].l; // (ML - SL)
}
occs[i].l = match_sa_cur.l; // Make SL = ML
break;
}
}
if(i < sams->n) {
// shadow
if(0 < k) {
//tmap_map1_aux_stack_shadow(k, bwt->seq_len, e->last_diff_offset, width_cur);
width_cur_i = seed2_len - (len - e->last_diff_offset);
tmap_map1_aux_stack_shadow(k, seed2_len, width_cur_i, width_cur);
}
sam_found = 0;
continue;
}
}
int32_t score = aln_score(e->n_mm, e->n_gapo, e->n_gape, opt);
int32_t do_add = 1;
if(sams->n == 0) {
best_score = score;
best_cnt = 0;
best_diff = e->n_mm + e->n_gapo + e->n_gape;
}
if(score == best_score) {
best_cnt += match_sa_cur.l - match_sa_cur.k + 1;
}
else {
if(best_diff + 1 <= max_diff) {
max_diff = best_diff + 1;
}
if(score < next_best_score) {
next_best_score = score;
}
else if(next_best_score < score) {
// no need to examine further
break;
}
}
if(do_add) { // append
uint32_t op, op_len, cigar_i;
tmap_map_sam_t *sam = NULL;
tmap_map1_aux_stack_entry_t *cur = NULL;
tmap_map_sams_realloc(sams, sams->n+1);
occs = tmap_realloc(occs, sizeof(tmap_map1_aux_occ_t) * sams->n, "occs");
sam = &sams->sams[sams->n-1];
sam->algo_id = TMAP_MAP_ALGO_MAP1;
sam->algo_stage = 0;
sam->score = e->score;
// aux data
tmap_map_sam_malloc_aux(sam);
k = occs[sams->n-1].k = match_sa_cur.k;
l = occs[sams->n-1].l= match_sa_cur.l;
sam->aux.map1_aux->n_mm = e->n_mm;
sam->aux.map1_aux->n_gapo = e->n_gapo;
sam->aux.map1_aux->n_gape = e->n_gape;
// aux data: reference length
cur = e;
i = e->i;
sam->aux.map1_aux->aln_ref = 0;
cigar_i = 0;
if(2 == sam_found) { // we used 'tmap_bwt_match_exact_alt_reverse'
op = STATE_M;
op_len = offset;
}
else {
op = -1;
op_len = 0;
}
while(0 <= i) {
cur = stack->entry_pool[i];
if(len == cur->offset) break;
if(op != cur->state) {
if(STATE_M == op || STATE_D == op) {
sam->aux.map1_aux->aln_ref += op_len;
}
op = cur->state;
op_len = 1;
}
else {
op_len++;
}
//fprintf(stderr, "cur->state=%c op_len=%d cur->prev_i=%d k=%u l=%u\n", "MIDS"[cur->state], op_len, cur->prev_i, cur->match_sa.k, cur->match_sa.l);
i = cur->prev_i;
}
if(STATE_M == op || STATE_D == op) {
sam->aux.map1_aux->aln_ref += op_len;
}
/*
fprintf(stderr, "shadow 2 k=%u l=%u len=%d offset=%d last_diff_offset=%d\n",
k, l, len, offset, e->last_diff_offset);
fprintf(stderr, "e->n_mm=%d e->n_gapo=%d e->n_gape=%d\n",
e->n_mm, e->n_gapo, e->n_gape);
*/
//tmap_map1_aux_stack_shadow(l - k + 1, bwt->seq_len, e->last_diff_offset, width_cur);
width_cur_i = seed2_len - (len - e->last_diff_offset);
tmap_map1_aux_stack_shadow(l - k + 1, seed2_len, width_cur_i, width_cur);
if(opt->max_best_cals < best_cnt) {
// ignore if too many "best" have been found
occs[sams->n-1].l -= (best_cnt - opt->max_best_cals); // only save the maximum
break;
}
}
}
else {
int32_t allow_diff = 1, allow_mm = (e->n_mm < max_mm) ? 1 : 0;
// decrement the offset
offset--;
// use a bound for mismatches
if(0 < offset) {
int32_t seed_width_cur_i = offset - (len - opt->seed_length);
width_cur_i = seed2_len - (len - offset);
if(0 < width_cur_i) {
if(m-1 < width_cur[width_cur_i-1].bid) {
allow_diff = 0;
}
else if(width_cur[width_cur_i-1].bid == m-1
&& width_cur[width_cur_i].bid == m-1
&& width_cur[width_cur_i-1].w == width_cur[width_cur_i].w) {
allow_mm = 0;
}
}
if(0 < seed_width_cur_i) {
if(NULL != seed_width_cur && 0 < seed_width_cur_i) {
if(n_seed_mm-1 < seed_width_cur[seed_width_cur_i-1].bid) {
allow_diff = 0;
}
else if(seed_width_cur[seed_width_cur_i-1].bid == n_seed_mm-1
&& seed_width_cur[seed_width_cur_i].bid == n_seed_mm-1
&& seed_width_cur[seed_width_cur_i-1].w == seed_width_cur[seed_width_cur_i].w) {
allow_mm = 0;
}
}
}
}
// retrieve the next SA interval
tmap_bwt_match_hash_2occ4(bwt, &e->match_sa, match_sa_next, hash);
// insertions/deletions
if(allow_diff
&& opt->indel_ends_bound + e->n_gapo + e->n_gape <= offset
&& opt->indel_ends_bound + e->n_gapo + e->n_gape <= len - offset) { // check to add gaps
if(STATE_M == e->state) { // gap open
if(e->n_gapo < max_gapo) { // gap open is allowed
// insertion
tmap_map1_aux_stack_push(stack, offset, &match_sa_cur, e->n_mm, e->n_gapo + 1, e->n_gape, STATE_I, 1, e, opt);
// deletion
for(j = 0; j != 4; ++j) {
if(match_sa_next[j].k <= match_sa_next[j].l) {
// remember that a gap deletion does not consume a
// read base, so use 'offset+1'
tmap_map1_aux_stack_push(stack, offset+1, &match_sa_next[j], e->n_mm, e->n_gapo + 1, e->n_gape, STATE_D, 1, e, opt);
}
}
}
}
else if(STATE_I == e->state) { // extension of an insertion
if(e->n_gape < max_gape) { // gap extension is allowed
tmap_map1_aux_stack_push(stack, offset, &match_sa_cur, e->n_mm, e->n_gapo, e->n_gape + 1, STATE_I, 1, e, opt);
}
}
else if(STATE_D == e->state) { // extension of a deletion
if(e->n_gape < max_gape) {
if(e->n_gape + e->n_gapo < max_diff || e->match_sa.l - e->match_sa.k + 1 < opt->max_cals_del) { // gap extension is allowed
for(j = 0; j != 4; ++j) {
if(match_sa_next[j].k <= match_sa_next[j].l) {
// remember that a gap deletion does not consume a
// read base, so use 'offset+1'
tmap_map1_aux_stack_push(stack, offset+1, &match_sa_next[j], e->n_mm, e->n_gapo, e->n_gape + 1, STATE_D, 1, e, opt);
}
}
}
}
}
}
// mismatches
if(1 == allow_mm && 1 == allow_diff) { // mismatches allowed
for(j=0;j<4;j++) {
int32_t c = (str[offset] + j) & 3;
int32_t is_mm = (0 < j || 3 < str[offset]);
if(match_sa_next[c].k <= match_sa_next[c].l) {
tmap_map1_aux_stack_push(stack, offset, &match_sa_next[c], e->n_mm + is_mm, e->n_gapo, e->n_gape, STATE_M, is_mm, e, opt);
}
}
}
else if(str[offset] < 4) { // try exact match only
int32_t c = str[offset] & 3;
if(match_sa_next[c].k <= match_sa_next[c].l) {
tmap_map1_aux_stack_push(stack, offset, &match_sa_next[c], e->n_mm, e->n_gapo, e->n_gape, STATE_M, 0, e, opt);
}
}
}
}
return tmap_map1_sam_to_real(sams, occs, bases, seed2_len, refseq, bwt, sa, hash, opt);
}
static inline void
tmap_map1_aux_stack_push(tmap_map1_aux_stack_t *stack,
int32_t offset,
tmap_bwt_match_occ_t *match_sa_prev,
int32_t n_mm, int32_t n_gapo, int32_t n_gape,
int32_t state, int32_t is_diff,
tmap_map1_aux_stack_entry_t *prev_entry,
const tmap_map_opt_t *opt)
{
int32_t i;
int32_t n_bins_needed = 0;
tmap_map1_aux_stack_entry_t *entry = NULL;
tmap_map1_aux_bin_t *bin = NULL;
// check to see if we need more memory
if(stack->entry_pool_length <= stack->entry_pool_i) {
int32_t i = stack->entry_pool_length;
stack->entry_pool_length <<= 2;
stack->entry_pool = tmap_realloc(stack->entry_pool,
sizeof(tmap_map1_aux_stack_entry_t*)*stack->entry_pool_length,
"stack->entry_pool");
while(i<stack->entry_pool_length) {
stack->entry_pool[i] = tmap_malloc(sizeof(tmap_map1_aux_stack_entry_t), "stack->entry_pool[i]");
i++;
}
}
entry = stack->entry_pool[stack->entry_pool_i];
entry->score = aln_score(n_mm, n_gapo, n_gape, opt);
entry->n_mm = n_mm;
entry->n_gapo = n_gapo;
entry->n_gape = n_gape;
entry->state = state;
entry->match_sa = (*match_sa_prev);
entry->i = stack->entry_pool_i;
entry->offset = offset;
if(NULL == prev_entry) {
entry->last_diff_offset = offset;
entry->prev_i = -1;
}
else {
entry->last_diff_offset = (1 == is_diff) ? (offset) : prev_entry->last_diff_offset;
entry->prev_i = prev_entry->i;
}
if(stack->n_bins <= entry->score) {
//tmap_bug();
// resize the bins if necessary
n_bins_needed = entry->score + 1;
// realloc
tmap_roundup32(n_bins_needed);
stack->bins = tmap_realloc(stack->bins, sizeof(tmap_map1_aux_bin_t) * n_bins_needed, "stack->bins");
// initialize
for(i=stack->n_bins;i<n_bins_needed;i++) {
stack->bins[i].n_entries = stack->bins[i].m_entries = 0;
stack->bins[i].entries = NULL;
}
stack->n_bins = n_bins_needed;
}
if(stack->n_bins <= entry->score) {
tmap_bug();
}
bin = &stack->bins[entry->score];
// - remove duplicates
// - most likely formed by tandem repeats or indels
// - too computationally expensive, and not necessary
/*
for(i=0;i<bin->n_entries;i++) {
if(bin->entries[i]->match_sa.k == entry->match_sa.k
&& bin->entries[i]->match_sa.l == entry->match_sa.l
&& bin->entries[i]->offset == entry->offset
&& bin->entries[i]->state == entry->state) {
return;
}
}
*/
// update best score
if(stack->best_score > entry->score) stack->best_score = entry->score;
if(bin->m_entries <= bin->n_entries) {
bin->m_entries++;
tmap_roundup32(bin->m_entries);
bin->entries = tmap_realloc(bin->entries, sizeof(tmap_map1_aux_bin_t) * bin->m_entries, "bin->entries");
}
bin->entries[bin->n_entries] = entry;
bin->n_entries++;
stack->entry_pool_i++;
stack->n_entries++;
}
gap_stack_t *gap_init_stack(int max_mm, int max_gapo, int max_gape, const gap_opt_t *opt)
{
return gap_init_stack2(aln_score(max_mm+1, max_gapo+1, max_gape+1, opt));
}
sint malign(sint istart,char *phylip_name) /* full progressive alignment*/
{
static sint *aligned;
static sint *group;
static sint ix;
sint *maxid, max, sum;
sint *tree_weight;
sint i,j,set,iseq=0;
sint status,entries;
lint score = 0;
info("Start of Multiple Alignment");
/* get the phylogenetic tree from *.ph */
if (nseqs >= 2)
{
status = read_tree(phylip_name, (sint)0, nseqs);
if (status == 0) return((sint)0);
}
/* calculate sequence weights according to branch lengths of the tree -
weights in global variable seq_weight normalised to sum to 100 */
calc_seq_weights((sint)0, nseqs, seq_weight);
/* recalculate tmat matrix as percent similarity matrix */
status = calc_similarities(nseqs);
if (status == 0) return((sint)0);
/* for each sequence, find the most closely related sequence */
maxid = (sint *)ckalloc( (nseqs+1) * sizeof (sint));
for (i=1;i<=nseqs;i++)
{
maxid[i] = -1;
for (j=1;j<=nseqs;j++)
if (j!=i && maxid[i] < tmat[i][j]) maxid[i] = tmat[i][j];
}
/* group the sequences according to their relative divergence */
if (istart == 0)
{
sets = (sint **) ckalloc( (nseqs+1) * sizeof (sint *) );
for(i=0;i<=nseqs;i++)
sets[i] = (sint *)ckalloc( (nseqs+1) * sizeof (sint) );
create_sets((sint)0,nseqs);
info("There are %d groups",(pint)nsets);
/* clear the memory used for the phylogenetic tree */
if (nseqs >= 2)
clear_tree(NULL);
/* start the multiple alignments......... */
info("Aligning...");
/* first pass, align closely related sequences first.... */
ix = 0;
aligned = (sint *)ckalloc( (nseqs+1) * sizeof (sint) );
for (i=0;i<=nseqs;i++) aligned[i] = 0;
for(set=1;set<=nsets;++set)
{
entries=0;
for (i=1;i<=nseqs;i++)
{
if ((sets[set][i] != 0) && (maxid[i] > divergence_cutoff))
{
entries++;
if (aligned[i] == 0)
{
if (output_order==INPUT)
{
++ix;
output_index[i] = i;
}
else output_index[++ix] = i;
aligned[i] = 1;
}
}
}
if(entries > 0) score = prfalign(sets[set], aligned);
else score=0.0;
/* negative score means fatal error... exit now! */
if (score < 0)
{
return(-1);
}
if ((entries > 0) && (score > 0))
info("Group %d: Sequences:%4d Score:%d",
(pint)set,(pint)entries,(pint)score);
else
info("Group %d: Delayed",
(pint)set);
}
for (i=0;i<=nseqs;i++)
sets[i]=ckfree((void *)sets[i]);
sets=ckfree(sets);
}
else
{
/* clear the memory used for the phylogenetic tree */
if (nseqs >= 2)
clear_tree(NULL);
aligned = (sint *)ckalloc( (nseqs+1) * sizeof (sint) );
ix = 0;
for (i=1;i<=istart+1;i++)
{
aligned[i] = 1;
++ix;
output_index[i] = i;
}
for (i=istart+2;i<=nseqs;i++) aligned[i] = 0;
}
/* second pass - align remaining, more divergent sequences..... */
/* if not all sequences were aligned, for each unaligned sequence,
find it's closest pair amongst the aligned sequences. */
group = (sint *)ckalloc( (nseqs+1) * sizeof (sint));
tree_weight = (sint *) ckalloc( (nseqs) * sizeof(sint) );
for (i=0;i<nseqs;i++)
tree_weight[i] = seq_weight[i];
/* if we haven't aligned any sequences, in the first pass - align the
two most closely related sequences now */
if(ix==0)
{
max = -1;
iseq = 0;
for (i=1;i<=nseqs;i++)
{
for (j=i+1;j<=nseqs;j++)
{
if (max < tmat[i][j])
{
max = tmat[i][j];
iseq = i;
}
}
}
aligned[iseq]=1;
if (output_order == INPUT)
{
++ix;
output_index[iseq] = iseq;
}
else
output_index[++ix] = iseq;
}
while (ix < nseqs)
{
for (i=1;i<=nseqs;i++) {
if (aligned[i] == 0)
{
maxid[i] = -1;
for (j=1;j<=nseqs;j++)
if ((maxid[i] < tmat[i][j]) && (aligned[j] != 0))
maxid[i] = tmat[i][j];
}
}
/* find the most closely related sequence to those already aligned */
max = -1;
iseq = 0;
for (i=1;i<=nseqs;i++)
{
if ((aligned[i] == 0) && (maxid[i] > max))
{
max = maxid[i];
iseq = i;
}
}
/* align this sequence to the existing alignment */
/* weight sequences with percent identity with profile*/
/* OR...., multiply sequence weights from tree by percent identity with new sequence */
if(no_weights==FALSE) {
for (j=0;j<nseqs;j++)
if (aligned[j+1] != 0)
seq_weight[j] = tree_weight[j] * tmat[j+1][iseq];
/*
Normalise the weights, such that the sum of the weights = INT_SCALE_FACTOR
*/
sum = 0;
for (j=0;j<nseqs;j++)
if (aligned[j+1] != 0)
sum += seq_weight[j];
if (sum == 0)
{
for (j=0;j<nseqs;j++)
seq_weight[j] = 1;
sum = j;
}
for (j=0;j<nseqs;j++)
if (aligned[j+1] != 0)
{
seq_weight[j] = (seq_weight[j] * INT_SCALE_FACTOR) / sum;
if (seq_weight[j] < 1) seq_weight[j] = 1;
}
}
entries = 0;
for (j=1;j<=nseqs;j++)
if (aligned[j] != 0)
{
group[j] = 1;
entries++;
}
else if (iseq==j)
{
group[j] = 2;
entries++;
}
aligned[iseq] = 1;
score = prfalign(group, aligned);
info("Sequence:%d Score:%d",(pint)iseq,(pint)score);
if (output_order == INPUT)
{
++ix;
output_index[iseq] = iseq;
}
else
output_index[++ix] = iseq;
}
group=ckfree((void *)group);
aligned=ckfree((void *)aligned);
maxid=ckfree((void *)maxid);
tree_weight=ckfree((void *)tree_weight);
aln_score();
/* make the rest (output stuff) into routine clustal_out in file amenu.c */
return(nseqs);
}
sint seqalign(sint istart,char *phylip_name) /* sequence alignment to existing profile */
{
static sint *aligned, *tree_weight;
static sint *group;
static sint ix;
sint *maxid, max;
sint i,j,status,iseq=0;
sint sum,entries;
lint score = 0;
info("Start of Multiple Alignment");
/* get the phylogenetic tree from *.ph */
if (nseqs >= 2)
{
status = read_tree(phylip_name, (sint)0, nseqs);
if (status == 0) return(0);
}
/* calculate sequence weights according to branch lengths of the tree -
weights in global variable seq_weight normalised to sum to 100 */
calc_seq_weights((sint)0, nseqs, seq_weight);
tree_weight = (sint *) ckalloc( (nseqs) * sizeof(sint) );
for (i=0;i<nseqs;i++)
tree_weight[i] = seq_weight[i];
/* recalculate tmat matrix as percent similarity matrix */
status = calc_similarities(nseqs);
if (status == 0) return((sint)0);
/* for each sequence, find the most closely related sequence */
maxid = (sint *)ckalloc( (nseqs+1) * sizeof (sint));
for (i=1;i<=nseqs;i++)
{
maxid[i] = -1;
for (j=1;j<=nseqs;j++)
if (maxid[i] < tmat[i][j]) maxid[i] = tmat[i][j];
}
/* clear the memory used for the phylogenetic tree */
if (nseqs >= 2)
clear_tree(NULL);
aligned = (sint *)ckalloc( (nseqs+1) * sizeof (sint) );
ix = 0;
for (i=1;i<=istart+1;i++)
{
aligned[i] = 1;
++ix;
output_index[i] = i;
}
for (i=istart+2;i<=nseqs;i++) aligned[i] = 0;
/* for each unaligned sequence, find it's closest pair amongst the
aligned sequences. */
group = (sint *)ckalloc( (nseqs+1) * sizeof (sint));
while (ix < nseqs)
{
if (ix > 0)
{
for (i=1;i<=nseqs;i++) {
if (aligned[i] == 0)
{
maxid[i] = -1;
for (j=1;j<=nseqs;j++)
if ((maxid[i] < tmat[i][j]) && (aligned[j] != 0))
maxid[i] = tmat[i][j];
}
}
}
/* find the most closely related sequence to those already aligned */
max = -1;
for (i=1;i<=nseqs;i++)
{
if ((aligned[i] == 0) && (maxid[i] > max))
{
max = maxid[i];
iseq = i;
}
}
/* align this sequence to the existing alignment */
entries = 0;
for (j=1;j<=nseqs;j++)
if (aligned[j] != 0)
{
group[j] = 1;
entries++;
}
else if (iseq==j)
{
group[j] = 2;
entries++;
}
aligned[iseq] = 1;
/* EITHER....., set sequence weights equal to percent identity with new sequence */
/*
for (j=0;j<nseqs;j++)
seq_weight[j] = tmat[j+1][iseq];
*/
/* OR...., multiply sequence weights from tree by percent identity with new sequence */
for (j=0;j<nseqs;j++)
seq_weight[j] = tree_weight[j] * tmat[j+1][iseq];
if (debug>1)
for (j=0;j<nseqs;j++) if (group[j+1] == 1)fprintf (stdout,"sequence %d: %d\n", j+1,tree_weight[j]);
/*
Normalise the weights, such that the sum of the weights = INT_SCALE_FACTOR
*/
sum = 0;
for (j=0;j<nseqs;j++)
if (group[j+1] == 1) sum += seq_weight[j];
if (sum == 0)
{
for (j=0;j<nseqs;j++)
seq_weight[j] = 1;
sum = j;
}
for (j=0;j<nseqs;j++)
{
seq_weight[j] = (seq_weight[j] * INT_SCALE_FACTOR) / sum;
if (seq_weight[j] < 1) seq_weight[j] = 1;
}
if (debug > 1) {
fprintf(stdout,"new weights\n");
for (j=0;j<nseqs;j++) if (group[j+1] == 1)fprintf( stdout,"sequence %d: %d\n", j+1,seq_weight[j]);
}
score = prfalign(group, aligned);
info("Sequence:%d Score:%d",(pint)iseq,(pint)score);
if (output_order == INPUT)
{
++ix;
output_index[iseq] = iseq;
}
else
output_index[++ix] = iseq;
}
group=ckfree((void *)group);
aligned=ckfree((void *)aligned);
maxid=ckfree((void *)maxid);
aln_score();
/* make the rest (output stuff) into routine clustal_out in file amenu.c */
return(nseqs);
}