/****************************************************************************
* Create a new alignment with any sequence that contains nothing but
* gap ('-') characters removed. Returns the new alignment. Does not
* change the old alignment.
* If there are no all-gap sequences, the returned alignment is the
* same object as the original alignment.
****************************************************************************/
static ALIGNMENT_T* remove_allgap_sequences(ALIGNMENT_T* alignment)
{
ALIGNMENT_T* new_alignment;
int i_aln;
int l_aln = get_num_aligned_sequences(alignment);
STRING_LIST_T* keeper_seqs = new_string_list();
// Identify the all-gap sequences.
for (i_aln=0; i_aln<l_aln; i_aln++) {
SEQ_T* sequence = get_alignment_sequence(i_aln, alignment);
int i_seq;
int l_seq = get_seq_length(sequence);
// Add sequence to keepers if it contains a non-gap.
for (i_seq=0; i_seq<l_seq; i_seq++) {
if (get_seq_char(i_seq, sequence) != '-') { // not gap?
add_string(get_seq_name(sequence), keeper_seqs); // non-gap: keeper
break;
}
}
}
// Remove any sequences not in keeper list.
if (get_num_strings(keeper_seqs) < l_aln) {
new_alignment = remove_alignment_seqs(keeper_seqs, alignment);
free_string_list(keeper_seqs);
} else {
new_alignment = alignment;
}
return(new_alignment);
} // remove_allgap_sequences
void assimilate_text_chars(string txt, Store &s, int order){
string chars;
queue<string> seq;
for (int i = 0; i < txt.length(); i++){
seq = get_seq_char(txt, i, order);
s.add_hits(seq, 1);
}
cerr << "ok" << endl;
}
/****************************************************************************
* Does a column of an alignment contain any ambiguity codes?
****************************************************************************/
BOOLEAN_T alignment_site_ambiguous(ALPH_T alph, int index, ALIGNMENT_T* alignment) {
int i;
for (i = 0; i < alignment->num_sequences; i++) {
char c = get_seq_char(index, alignment->sequences[i]);
if (!alph_is_concrete(alph, c)) return(TRUE);
}
return(FALSE);
}
/****************************************************************************
* Fill in a null terminated string with the bases in one column of the
* alignment. The user must allocate the memory for the string, which should
* be large enough to store one characters from each sequence in the alignment
* plus the trailing null character. This is done for reasons of efficiency,
* since in most cases the user will be making for this call iteratively over
* the length of the alignment.
****************************************************************************/
void get_alignment_col(int col, char* alignment_col, ALIGNMENT_T* alignment) {
int i;
for (i = 0; i < alignment->num_sequences; i++, alignment_col++) {
*alignment_col = get_seq_char(col, alignment->sequences[i]);
}
*alignment_col = '\0';
}
/****************************************************************************
* Does a column of an alignment contain gaps?
****************************************************************************/
BOOLEAN_T alignment_site_has_gaps(int index, ALIGNMENT_T* alignment) {
int i = 0;
for (i = 0; i < alignment->num_sequences; i++) {
char c = get_seq_char(index, alignment->sequences[i]);
if (c == '-' || c == '.') {
return(TRUE);
}
}
return(FALSE);
}
/****************************************************************************
* Get a cumulative count of gaps within one sequence of the alignment
****************************************************************************/
int* get_cumulative_gap_count(int seqIndex, ALIGNMENT_T* alignment) {
int* results = (int *)mm_malloc( sizeof(int) * alignment->length );
int i = 0;
char c = get_seq_char(i, alignment->sequences[seqIndex]);
if (c == '-' || c == '.') {
results[i] = 1;
} else {
results[i] = 0;
}
for (i = 1; i < alignment->length; i++) {
c = get_seq_char(i, alignment->sequences[seqIndex]);
if (c == '-' || c == '.') {
results[i] = results[i-1] + 1;
} else {
results[i] = results[i-1];
}
}
return results;
}
SEQ_T* get_consensus_sequence(double threshold, ALIGNMENT_T* alignment) {
char* seq_string = NULL;
unsigned char c = 0;
unsigned char most_freq_char = 0;
#define NUM_CHARS 127
char char_counts[NUM_CHARS];
int i = 0;
int j = 0;
double max_char_freq = 0.0;
SEQ_T* consensus;
assert(alignment != NULL);
seq_string = mm_malloc(alignment->length * sizeof(char) + 1);
if (seq_string == NULL) {
die("Error allocating consensus sequence string\n");
}
// For each column in the alignment
for (i = 0; i < alignment->length; i++) {
most_freq_char = 0;
memset(char_counts, 0, NUM_CHARS * sizeof(char));
// Count character occurances
for (j = 0; j < alignment->num_sequences; j++) {
c = get_seq_char(i, alignment->sequences[j]);
char_counts[c]++;
}
// Find the index of the character that occurs the most frequently
for (c = 0; c < NUM_CHARS; c++) {
most_freq_char = char_counts[most_freq_char] >= char_counts[c] ?
most_freq_char : c;
}
// If the most frequent character exceeds the threshold
// it will be the consensus character.
max_char_freq = (double) char_counts[most_freq_char] /
(double) alignment->num_sequences;
if (max_char_freq >= threshold) {
seq_string[i] = most_freq_char;
} else {
// Otherwise the consensus is the gap character
seq_string[i] = '-';
}
}
seq_string[i] = '\0';
consensus = allocate_seq("Consensus", "", 0, seq_string);
if (seq_string != NULL) myfree(seq_string);
return(consensus);
}
/****************************************************************************
* Return an array containing the frequencies in the alignment for each
* character of the alphabet. Gaps and ambiguity characters other then
* ANY_BASE are not counted. The freq. of ANY_BASE characters is stored
* in the last element of the array.
****************************************************************************/
ARRAY_T* get_alignment_freqs(ALPH_T alph, ALIGNMENT_T* alignment) {
char c = 0;
int aindex = 0;
int asize = 0;
int i = 0;
int s = 0;
int total_bases = 0;
int* num_bases = NULL;
ARRAY_T* freqs = NULL;
// Initialize counts for each character in the alphabet
asize = alph_size(alph, ALPH_SIZE);
num_bases = mm_malloc(asize * sizeof(int));
for (i = 0; i < asize; i++) {
num_bases[i] = 0;
}
for (s = 0; s < alignment->num_sequences; s++) {
for (i = 0; i < alignment->length; i++) {
c = get_seq_char(i, alignment->sequences[s]);
aindex = alph_index(alph, c);
// c might be an ambiguity code. We don't count ambiguity codes.
if (aindex != -1 && aindex < asize) {
num_bases[aindex]++;
total_bases++;
}
}
}
freqs = allocate_array(asize);
for (i = 0; i < asize; i++) {
set_array_item(i, (double) num_bases[i] / (double) total_bases, freqs);
}
// Clean up the count of characters
myfree(num_bases);
return freqs;
}
/****************************************************************************
* Return an array containing the frequencies in the sequence for each
* character of the alphabet. Characters not in the alphabet are not
* counted.
****************************************************************************/
ARRAY_T* get_sequence_freqs
(SEQ_T* seq, ALPH_T alph)
{
char c = 0;
int a_index = 0;
int a_size = 0;
int i = 0;
int total_bases = 0;
int* num_bases = NULL;
ARRAY_T* freqs = NULL;
// Initialize counts for each character in the alphabet
a_size = alph_size(alph, ALPH_SIZE);
num_bases = mm_malloc(a_size * sizeof(int));
for (i = 0; i < a_size; i++) {
num_bases[i] = 0;
}
for (i = 0; i < seq->length; i++) {
c = get_seq_char(i, seq);
if (c != '-' && c != '-') {
a_index = alph_index(alph, c);
if (a_index == -1 || a_index >= a_size) continue;
num_bases[a_index]++;
total_bases++;
}
}
freqs = allocate_array(a_size);
for (i = 0; i < a_size; i++) {
set_array_item(i, (double) num_bases[i] / (double) total_bases, freqs);
}
// Clean up the count of characters
myfree(num_bases);
return freqs;
}
/****************************************************************************
* Return an array containing the frequencies in the sequences for each
* character of the alphabet. Characters not in the alphabet are not
* counted.
*
* When seq is provided it returns null, otherwise it converts the accumulated
* result in bgcalc into a background.
*
*
* Pseudocode example:
* ALPH_T alph = ...
* BGCALC_T *bgcalc = NULL;
* for each seq:
* calculate_background(alph, seq, &bgcalc);
* ARRAY_T *bg = calculate_background(NULL, &bgcalc);
****************************************************************************/
ARRAY_T* calculate_background(
ALPH_T alph,
SEQ_T* seq,
BGCALC_T** bgcalc
){
BGCALC_T *calc;
int a_size, i, a_index;
char c;
double freq, chunk_part, chunk_freq;
ARRAY_T *background;
assert(bgcalc != NULL);
assert(seq != NULL || *bgcalc != NULL);
// get the alphabet
// get the alphabet size
a_size = alph_size(alph, ALPH_SIZE);
if (*bgcalc == NULL) {
//allocate and initialize calc
calc = mm_malloc(sizeof(BGCALC_T));
calc->alph = alph;
calc->chunk_seen = 0;
calc->weight = 0;
calc->chunk_counts = mm_malloc(a_size * sizeof(long));
calc->bg = mm_malloc(a_size * sizeof(double));
for (i = 0; i < a_size; ++i) {
calc->chunk_counts[i] = 0;
calc->bg[i] = 0;
}
*bgcalc = calc;
} else {
calc = *bgcalc;
assert(alph == calc->alph);
if (calc->weight == LONG_MAX) return NULL;
}
if (seq == NULL) {
// no sequence so calculate the final result
background = allocate_array(alph_size(alph, ALL_SIZE));
if (calc->weight == 0) {
if (calc->chunk_seen > 0) {
// when we haven't had to approximate yet
// just do a normal background calculation
for (i = 0; i < a_size; i++) {
freq = (double) calc->chunk_counts[i] / (double) calc->chunk_seen;
set_array_item(i, freq, background);
}
} else {
fputs("Uniform\n", stdout);
// when there are no counts then return uniform
freq = (double) 1 / (double) a_size;
for (i = 0; i < a_size; i++) {
set_array_item(i, freq, background);
}
}
} else {
if (calc->chunk_seen > 0) {
// combine the frequencies for the existing chunks with the counts
// for the partially completed chunk
chunk_part = (double) calc->chunk_seen / (double) BG_CALC_CHUNK;
for (i = 0; i < a_size; i++) {
chunk_freq = (double) calc->chunk_counts[i] /
(double) calc->chunk_seen;
freq = ((calc->bg[i] * calc->weight) + (chunk_freq * chunk_part)) /
(calc->weight + chunk_part);
set_array_item(i, freq, background);
}
} else {
// in the odd case we get to an integer number of chunks
for (i = 0; i < a_size; i++) {
set_array_item(i, calc->bg[i], background);
}
}
}
calc_ambigs(alph, FALSE, background);
// free bgcalc structure
free(calc->bg);
free(calc->chunk_counts);
free(calc);
*bgcalc = NULL;
return background;
}
// we have a sequence to add to the background calculation
for (i = 0; i < seq->length; i++) {
c = get_seq_char(i, seq);
a_index = alph_index(alph, c);
if (a_index == -1 || a_index >= a_size) continue;
calc->chunk_counts[a_index]++;
calc->chunk_seen++;
if (calc->chunk_seen == BG_CALC_CHUNK) {
if (calc->weight == 0) {
for (i = 0; i < a_size; i++) {
calc->bg[i] = (double) calc->chunk_counts[i] / (double) BG_CALC_CHUNK;
}
} else {
for (i = 0; i < a_size; i++) {
chunk_freq = (double) calc->chunk_counts[i] / (double) BG_CALC_CHUNK;
calc->bg[i] = (calc->bg[i] * calc->weight + chunk_freq) /
(calc->weight + 1);
}
}
calc->weight++;
// reset the counts for the next chunk
for (i = 0; i < a_size; i++) {
calc->chunk_counts[i] = 0;
}
calc->chunk_seen = 0;
// I don't think it is feasible to reach this limit
// but I guess I'd better check anyway
if (calc->weight == LONG_MAX) {
fprintf(stderr, "Sequence data set is so large that even the "
"approximation designed for large datasets can't handle it!");
return NULL;
}
}
}
return NULL;
}
/****************************************************************************
* Remove from the alignment all columns that contain gaps for the
* specified species.
****************************************************************************/
ALIGNMENT_T* remove_alignment_gaps
(char* species,
ALIGNMENT_T* alignment)
{
// Locate this species in the alignment.
int species_index = get_index_in_string_list(species,
get_species_names(alignment));
if (species_index == -1) {
die("Can't find %s in alignment.\n", species);
}
SEQ_T* this_seq = get_alignment_sequence(species_index, alignment);
// Get the dimensions of the original matrix.
int num_sequences = get_num_aligned_sequences(alignment);
int alignment_length = get_alignment_length(alignment);
// Allocate memory for raw sequences that will constitute the new alignment.
char** raw_sequences = (char**)mm_malloc(sizeof(char*) * num_sequences);
int i_seq = 0;
for (i_seq = 0; i_seq < num_sequences; i_seq++) {
raw_sequences[i_seq]
= (char*)mm_calloc(alignment_length + 1, sizeof(char*));
}
char* consensus = get_consensus_string(alignment);
char* new_consensus
= (char*)mm_calloc(alignment_length + 1, sizeof(char*));
// Iterate over all columns.
int i_column;
int i_raw = 0;
for (i_column = 0; i_column < alignment_length; i_column++) {
// Is there a gap?
char this_char = get_seq_char(i_column, this_seq);
if ((this_char != '-') && (this_char != '.')) {
// If no gap, then copy over this column.
for (i_seq = 0; i_seq < num_sequences; i_seq++) {
SEQ_T* this_sequence = get_alignment_sequence(i_seq, alignment);
char this_char = get_seq_char(i_column, this_sequence);
raw_sequences[i_seq][i_raw] = this_char;
}
new_consensus[i_raw] = consensus[i_column];
i_raw++;
}
}
// Create new sequence objects.
SEQ_T** new_sequences = (SEQ_T**)mm_malloc(num_sequences * sizeof(SEQ_T*));
for (i_seq = 0; i_seq < num_sequences; i_seq++) {
SEQ_T* this_sequence = get_alignment_sequence(i_seq, alignment);
new_sequences[i_seq] = allocate_seq(get_seq_name(this_sequence),
get_seq_description(this_sequence),
get_seq_offset(this_sequence),
raw_sequences[i_seq]);
}
// Allocate and return the new alignment.
ALIGNMENT_T* new_alignment
= allocate_alignment(get_alignment_name(alignment),
get_alignment_description(alignment),
num_sequences,
new_sequences,
new_consensus);
// Free local dynamic memory.
for (i_seq = 0; i_seq < num_sequences; i_seq++) {
myfree(raw_sequences[i_seq]);
free_seq(new_sequences[i_seq]);
}
myfree(raw_sequences);
myfree(new_sequences);
myfree(new_consensus);
return(new_alignment);
}
