/****************************************************************************
* Remove from an alignment any sequence whose ID is not in a given list.
*
* N.B. It is NOT an error for the given list to contain sequence IDs that
* are not in the alignment.
****************************************************************************/
ALIGNMENT_T* remove_alignment_seqs
(STRING_LIST_T* seqs_to_keep,
ALIGNMENT_T* alignment)
{
// Extract the names of the sequences in the alignment.
STRING_LIST_T* alignment_species = get_species_names(alignment);
int num_species = get_num_strings(alignment_species);
// Count how many sequences will be in the new alignment.
int i_species;
int num_final = 0;
for (i_species = 0; i_species < num_species; i_species++) {
char* this_species = get_nth_string(i_species, alignment_species);
if (have_string(this_species, seqs_to_keep)) {
num_final++;
} else {
if (verbosity >= NORMAL_VERBOSE) {
fprintf(stderr, "Removing %s from alignment.\n", this_species);
}
}
}
// Allocate space for the new sequences.
SEQ_T** new_sequences = (SEQ_T**)mm_malloc(num_final * sizeof(SEQ_T*));
// Copy the sequences.
int final_index = 0;
num_species = get_num_strings(seqs_to_keep);
for (i_species = 0; i_species < num_species; i_species++) {
char* this_species = get_nth_string(i_species, seqs_to_keep);
// If the requested ID is in the alignment, then copy over the sequence.
if (have_string(this_species, alignment_species)) {
SEQ_T* this_seq
= get_alignment_sequence_by_name(this_species, alignment);
new_sequences[final_index] = copy_sequence(this_seq);
final_index++;
}
}
// Allocate and return the new alignment.
char *consensus = NULL;
copy_string(&consensus, get_consensus_string(alignment));
ALIGNMENT_T* new_alignment
= allocate_alignment(get_alignment_name(alignment),
get_alignment_description(alignment),
num_final,
new_sequences,
consensus);
return(new_alignment);
}
MATRIX_T *read_score_matrix(
char *score_filename, /* name of score file */
char **alpha1 /* alphabet in score matrix */
)
{
int i;
char *alpha; /* alphabet in file */
int alen;
FILE *score_file;
RDB_MATRIX_T *rdb_matrix;
/* open the score file */
if (open_file(score_filename, "r", FALSE, "score", "substitution scores", &score_file) == 0)
exit(1);
/* read in the score file */
rdb_matrix = read_rdb_matrix(" ", FALSE, 0, FALSE, NULL, score_file);
/* get alphabet */
alen = get_num_strings(rdb_matrix->col_names);
alpha = (char *)mm_malloc(sizeof(char) * (alen+1));
for (i=0; i<alen; i++) alpha[i] = get_nth_string(i, rdb_matrix->col_names)[0];
alpha[i] = '\0';
*alpha1 = alpha; /* return alphabet */
return(rdb_matrix->matrix);
} /* read_score_matrix */
/****************************************************************************
* 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 ramen_get_scores() {
int i;
int seq_num;
seq_num = get_num_strings(seq_ids); //number of sequences
//allocate space for final one result per motif array.
rsr = malloc(sizeof(ramen_result_t*)*motifs.num);
for(i=0;i<motifs.num;i++) {
fprintf(stderr, "\rScoring %i of %i motifs...", i+1, motifs.num);
rsr[i] = ramen_do_linreg_test(i);
}
//Order by MSE.
qsort(rsr, motifs.num, sizeof(ramen_result_t*), ramen_compare_mse);
fprintf(stderr, "\n");
}
/***********************************************************************
* Should the given motif be inserted into the model?
* FIXME: These tests needn't be mutually exclusive.
***********************************************************************/
static BOOLEAN_T retain_motif(
STRING_LIST_T* requested_motifs, // IDs of motifs to include.
double e_threshold, // E-value to include motifs.
double complexity_threshold, // Complexity threshold to include.
ORDER_T* order_spacing, // Motif order and spacing (linear HMM).
MOTIF_T* motif // The motif.
) {
int num_requested;
int i;
char* motif_id;
/* Method 1: Select motifs by index. */
num_requested = get_num_strings(requested_motifs);
if (num_requested > 0) {
motif_id = get_motif_id(motif);
for (i = 0; i < num_requested; i++) {
if (strcmp(get_nth_string(i, requested_motifs), motif_id) == 0) {
return(TRUE);
}
}
return(FALSE);
}
/* Method 2: Select motifs below a certain E-value threshold. */
else if (e_threshold != 0.0) {
return (get_motif_evalue(motif) <= e_threshold);
}
/* Method 3: Select motifs that are included in the order string. */
else if (order_spacing != NULL) {
return order_contains(get_motif_id(motif), order_spacing);
}
// Method 4: Select motifs by their complexity score.
else if (complexity_threshold != 0.0) {
return(motif->complexity >= complexity_threshold);
}
/* Default is to include all motifs. */
return(TRUE);
}
/*************************************************************************
* int main
*************************************************************************/
int main(int argc, char *argv[])
{
/* Data structures. */
int num_motifs; /* The number of motifs in the model. */
MOTIF_T motifs[2 * MAX_MOTIFS]; /* The motifs. */
STRING_LIST_T* motif_occurrences = NULL; /* Strings describing occurrences of
motifs */
BOOLEAN_T has_reverse_strand = FALSE; /* MEME file contained both strands */
ARRAY_T* background; /* Background probs for alphabet. */
ORDER_T* order_spacing; /* Linear HMM order and spacing. */
MATRIX_T* transp_freq = NULL; /* Matrix of inter-motif transitions freqs. */
MATRIX_T* spacer_ave = NULL; /* Matrix of average spacer lengths. */
MHMM_T * the_hmm = NULL; /* The HMM being constructed. */
/* Command line parameters. */
char * meme_filename; /* Input file containg motifs. */
char * hmm_type_str; /* HMM type. */
HMM_T hmm_type;
STRING_LIST_T* requested_motifs; /* Indices of requested motifs. */
int request_n; /* The user asked for the first n motifs. */
double e_threshold; /* E-value threshold for motif inclusion. */
double complexity_threshold; // For eliminating low-complexity motifs.
double p_threshold; /* p-value threshold for motif occurences. */
char* order_string; /* Motif order and spacing. */
int spacer_states; /* Number of states in each spacer. */
BOOLEAN_T fim; /* Represent spacers as free insertion
modules? */
BOOLEAN_T keep_unused; // Drop unused inter-motif transitions?
double trans_pseudo; /* Transition pseudocount. */
double spacer_pseudo; // Spacer (self-loop) pseudocount. */
char* description; // Descriptive text to be stored in model.
BOOLEAN_T print_header; /* Print file header? */
BOOLEAN_T print_params; /* Print parameter summary? */
BOOLEAN_T print_time; /* Print timing data (dummy: always false). */
/* Local variables. */
int i_motif;
/**********************************************
* COMMAND LINE PROCESSING
**********************************************/
// Define command line options.
cmdoption const options[] = {
{"type", OPTIONAL_VALUE},
{"description", REQUIRED_VALUE},
{"motif", REQUIRED_VALUE},
{"nmotifs", REQUIRED_VALUE},
{"ethresh", REQUIRED_VALUE},
{"lowcomp", REQUIRED_VALUE},
{"pthresh", REQUIRED_VALUE},
{"order", REQUIRED_VALUE},
{"nspacer", REQUIRED_VALUE},
{"fim", NO_VALUE},
{"keep-unused", NO_VALUE},
{"transpseudo", REQUIRED_VALUE},
{"spacerpseudo", REQUIRED_VALUE},
{"verbosity", REQUIRED_VALUE},
{"noheader", NO_VALUE},
{"noparams", NO_VALUE},
{"notime", NO_VALUE},
{"quiet", NO_VALUE},
};
int option_count = 18;
int option_index = 0;
// Define the usage message.
char usage[1000] = "";
strcat(usage, "USAGE: mhmm [options] <MEME file>\n");
strcat(usage, "\n");
strcat(usage, " Options:\n");
strcat(usage, " --type [linear|complete|star] (default=linear)\n");
strcat(usage, " --description <string> (may be repeated)\n");
strcat(usage, " --motif <motif #> (may be repeated)\n");
strcat(usage, " --nmotifs <#>\n");
strcat(usage, " --ethresh <E-value>\n");
strcat(usage, " --lowcomp <value>\n");
strcat(usage, " --pthresh <p-value>\n");
strcat(usage, " --order <string>\n");
strcat(usage, " --nspacer <spacer length> (default=1)\n");
strcat(usage, " --fim\n");
strcat(usage, " --keep-unused\n");
strcat(usage, " --transpseudo <pseudocount>\n");
strcat(usage, " --spacerpseudo <pseudocount>\n");
strcat(usage, " --verbosity 1|2|3|4|5 (default=2)\n");
strcat(usage, " --noheader\n");
strcat(usage, " --noparams\n");
strcat(usage, " --notime\n");
strcat(usage, " --quiet\n");
strcat(usage, "\n");
/* Make sure various options are set to NULL or defaults. */
meme_filename = NULL;
hmm_type_str = NULL;
hmm_type = INVALID_HMM;
requested_motifs = new_string_list();
request_n = 0;
e_threshold = 0.0;
complexity_threshold = 0.0;
p_threshold = 0.0;
order_string = NULL;
spacer_states = DEFAULT_SPACER_STATES,
fim = FALSE;
keep_unused = FALSE;
trans_pseudo = DEFAULT_TRANS_PSEUDO;
spacer_pseudo = DEFAULT_SPACER_PSEUDO;
description = NULL;
print_header = TRUE;
print_params = TRUE;
print_time = FALSE;
simple_setopt(argc, argv, option_count, options);
// Parse the command line.
while (1) {
int c = 0;
char* option_name = NULL;
char* option_value = NULL;
const char * message = NULL;
// Read the next option, and break if we're done.
c = simple_getopt(&option_name, &option_value, &option_index);
if (c == 0) {
break;
} else if (c < 0) {
simple_getopterror(&message);
die("Error processing command line options (%s)\n", message);
}
if (strcmp(option_name, "type") == 0) {
if (option_value != NULL) {
hmm_type_str = option_value;
}
} else if (strcmp(option_name, "description") == 0) {
description = option_value;
} else if (strcmp(option_name, "motif") == 0) {
add_string(option_value, requested_motifs);
} else if (strcmp(option_name, "nmotifs") == 0) {
request_n = atoi(option_value);
} else if (strcmp(option_name, "ethresh") == 0) {
e_threshold = atof(option_value);
} else if (strcmp(option_name, "lowcomp") == 0) {
complexity_threshold = atof(option_value);
} else if (strcmp(option_name, "pthresh") == 0) {
p_threshold = atof(option_value);
} else if (strcmp(option_name, "order") == 0) {
order_string = option_value;
} else if (strcmp(option_name, "nspacer") == 0) {
spacer_states = atoi(option_value);
} else if (strcmp(option_name, "fim") == 0) {
fim = TRUE;
} else if (strcmp(option_name, "keep-unused") == 0) {
keep_unused = TRUE;
} else if (strcmp(option_name, "transpseudo") == 0) {
trans_pseudo = atof(option_value);
} else if (strcmp(option_name, "spacerpseudo") == 0) {
spacer_pseudo = atof(option_value);
} else if (strcmp(option_name, "verbosity") == 0) {
verbosity = (VERBOSE_T)atoi(option_value);
} else if (strcmp(option_name, "noheader") == 0) {
print_header = FALSE;
} else if (strcmp(option_name, "noparams") == 0) {
print_params = FALSE;
} else if (strcmp(option_name, "notime") == 0) {
print_time = FALSE;
} else if (strcmp(option_name, "quiet") == 0) {
print_header = print_params = print_time = FALSE;
verbosity = QUIET_VERBOSE;
}
}
// Read the single required argument.
if (option_index + 1 != argc) {
fprintf(stderr, "%s", usage);
exit(1);
}
meme_filename = argv[option_index];
// Set up motif requests.
if (request_n != 0) {
if (get_num_strings(requested_motifs) != 0) {
die("Can't combine the -motif and -nmotifs options.\n");
} else {
for (i_motif = 0; i_motif < request_n; i_motif++) {
char motif_id[MAX_MOTIF_ID_LENGTH + 1];
sprintf(motif_id, "%d", i_motif + 1);
add_string(motif_id, requested_motifs);
}
}
}
/* Set the model type. */
hmm_type = convert_enum_type_str(hmm_type_str, LINEAR_HMM, HMM_STRS,
NUM_HMM_T);
/* Gotta have positive spacer length. */
if (spacer_states <= 0) {
die("Negative spacer length (%d).\n", spacer_states);
}
/* Make sure motifs weren't selected redundantly. */
// FIXME: Add tests for complexity threshold.
if ((get_num_strings(requested_motifs) != 0) && (e_threshold != 0.0)) {
die("Can't use -motif or -nmotifs with -ethresh.");
}
if ((get_num_strings(requested_motifs) != 0) && (order_string != NULL)) {
die("Can't use -motif or -nmotifs with -order.");
}
if ((order_string != NULL) && (e_threshold != 0.0)) {
die("Can't use -ethresh and -order.");
}
/* Prevent trying to build a complete or star model with ordering. */
if (order_string != NULL) {
if (hmm_type == COMPLETE_HMM)
die("Can't specify motif order with a completely connected model.");
else if (hmm_type == STAR_HMM)
die("Can't specify motif order with a star model.");
}
// Parse the order string.
order_spacing = create_order(order_string);
/**********************************************
* READING THE MOTIFS
**********************************************/
BOOLEAN_T read_file = FALSE;
double pseudocount = 0;
read_meme_file(
meme_filename,
"motif-file", // Take bg freq. from motif file.
pseudocount,
REQUIRE_PSPM,
&num_motifs,
motifs,
&motif_occurrences,
&has_reverse_strand,
&background
);
process_raw_motifs_for_model(
&num_motifs,
motifs,
motif_occurrences,
requested_motifs,
has_reverse_strand,
keep_unused,
p_threshold,
e_threshold,
complexity_threshold,
&order_spacing,
&transp_freq,
&spacer_ave,
trans_pseudo,
spacer_pseudo
);
/**********************************************
* BUILDING THE HMM
**********************************************/
/* Build the motif-based HMM. */
if (hmm_type == LINEAR_HMM) {
if (order_spacing != NULL) {
reorder_motifs(order_spacing, &num_motifs, motifs);
}
else {
die("No order specified for the motifs.\n"
"For the linear model the motif file must contain motif occurence\n"
"data or the motif order must be specified using "
"the --order option.");
}
build_linear_hmm(
background,
order_spacing,
spacer_states,
motifs,
num_motifs,
fim,
&the_hmm
);
} else if (hmm_type == COMPLETE_HMM) {
build_complete_hmm(
background,
spacer_states,
motifs,
num_motifs,
transp_freq,
spacer_ave,
fim,
&the_hmm
);
} else if (hmm_type == STAR_HMM) {
build_star_hmm(
background,
spacer_states,
motifs,
num_motifs,
fim,
&the_hmm
);
}
// Add some global information.
copy_string(&(the_hmm->motif_file), meme_filename);
/**********************************************
* WRITING THE HMM
**********************************************/
/* Print the header. */
if (print_header)
write_header(
program,
"",
description,
meme_filename,
NULL,
NULL,
stdout
);
/* Write the HMM. */
write_mhmm(verbosity, the_hmm, stdout);
/* Print the program parameters. */
if (print_params) {
printf("Program parameters for mhmm\n");
printf(" MEME file: %s\n", meme_filename);
printf(" Motifs:");
write_string_list(" ", requested_motifs, stdout);
printf("\n");
printf(" Model topology: %s\n",
convert_enum_type(hmm_type, HMM_STRS, NUM_HMM_T));
printf(" States per spacer: %d\n", spacer_states);
printf(" Spacers are free-insertion modules: %s\n",
boolean_to_string(fim));
printf("\n");
}
free_array(background);
free_string_list(requested_motifs);
free_order(order_spacing);
free_matrix(transp_freq);
free_matrix(spacer_ave);
for (i_motif = 0; i_motif < num_motifs; i_motif++)
free_motif(&(motifs[i_motif]));
free_mhmm(the_hmm);
return(0);
}
/***********************************************************************
* Select the motifs used to build the model, parse any motif
* occurences, build the motif order object, and the motif
* and spacer frequency matrices.
***********************************************************************/
void process_raw_motifs_for_model(
int* num_motifs, // Number of motifs. IN, OUT
MOTIF_T* motifs, // Array of motifs IN, OUT
STRING_LIST_T* motif_occurrences, // List of motif occurrences. OUT
STRING_LIST_T* requested_motifs, // Explicitly requested motifs. IN
BOOLEAN_T has_reverse_strand, // Did file contain both strands? IN
BOOLEAN_T keep_unused, // Retain unsed motifs? IN
double p_threshold, // Motif p-value threshold IN
double e_threshold, // Motif e-value threshold IN
double complexity_threshold, // Motif complexity threshold IN
ORDER_T** order_spacing, // Motif/spacer order IN, OUT
MATRIX_T** transp_freq, // Motif transition freqs OUT
MATRIX_T** spacer_ave, // Spacer transition freqs OUT
double trans_pseudo, // Motif transition pseudo-counts IN
double spacer_pseudo // Spacer transition pseudo-counts IN
) {
// If both strands, make reverse complements.
if (has_reverse_strand) {
add_reverse_complements(num_motifs, motifs);
}
/* Remove motifs not allowed by the command line parameters */
filter_motifs(
requested_motifs,
e_threshold,
complexity_threshold,
order_spacing,
num_motifs,
motifs
);
/* Turn the raw motifs and motif occurences into the */
/* elements of the model */
if (motif_occurrences != NULL && get_num_strings(motif_occurrences) > 0) {
parse_motif_occurrences(
motif_occurrences,
has_reverse_strand,
p_threshold,
order_spacing,
transp_freq,
spacer_ave,
*num_motifs,
motifs
);
}
else {
// If no occurrences are found, initialize matrices uniformly.
compute_naive_transitions_and_spacers(
*num_motifs,
transp_freq,
spacer_ave
);
}
// Convert spacer info to probabilities.
normalize_spacer_counts(
trans_pseudo,
spacer_pseudo,
keep_unused,
*transp_freq,
*spacer_ave);
// Throw out unused motifs.
throw_out_unused_motifs(*transp_freq, *spacer_ave, num_motifs, motifs);
}
/***********************************************************************
* Parse the motif occurrences.
*
* Each motif occurence string contains the following items
* - sequence id,
* - sequence p-value,
* - number n of motif occurrences, and
* - length of sequence.
*
* This is followed by n triples containing
* - motif id,
* - occurrence position, and
* - occurrence p-value.
*
***********************************************************************/
static void parse_motif_occurrences(
STRING_LIST_T* motif_occurrences, // List of motif occurences OUT
BOOLEAN_T has_reverse_strand, // File included both strands? IN
double p_threshold, // P-value to include motif occurences. OUT
ORDER_T** order_spacing, // Motif order and spacing (linear HMM)
// IN OUT.
MATRIX_T** transp_freq, // Motif-to-motif transitions. OUT
MATRIX_T** spacer_ave, // Average inter-motif distances. OUT
int num_motifs, // Number of motifs retrieved. IN
MOTIF_T* motifs // The retrieved motifs. IN
) {
ORDER_T* new_order; // New order and spacing.
BOOLEAN_T find_order; // Should we look for the motif order?
// If we already have a motif order and spacing, don't find any more.
if (*order_spacing == NULL) {
find_order = TRUE;
} else {
find_order = FALSE;
}
new_order = NULL;
// Allocate the matrices.
*transp_freq = allocate_matrix(num_motifs + 2, num_motifs + 2);
*spacer_ave = allocate_matrix(num_motifs + 2, num_motifs + 2);
init_matrix(0.0, *transp_freq);
init_matrix(0.0, *spacer_ave);
int num_occurrence_strings = get_num_strings(motif_occurrences);
int i;
for (i = 0; i < num_occurrence_strings; i++) {
char* sequence_id; // ID of the current sequence.
float sequence_p; // pvalue of the entire sequence.
int num_occurs; // Number of motif occurences in this sequence.
int seq_length; // Length of the current sequence.
int i_occur; // Index of the current occurrence.
char prev_motif[MAX_MOTIF_ID_LENGTH + 1]; // Index of the previous motif.
int prev_position; // Location of the right edge of previous motif.
float motif_p; // P-value of the current occurrence.
char *c; // Dummy to hold return of strtok.
char* line = get_nth_string(i, motif_occurrences);
/* Read the sequence identifier, p-value, number of occurrences
and length. */
// tlb; sscanf crashes if strtok returns NULL so pass it "" then
sequence_id = strtok(line, " ");
if (sequence_id == NULL) {
die("Error reading motif occurrences.\n%s", line);
}
if (sscanf((c=strtok(NULL, " "))?c:"", "%f", &sequence_p) != 1) {
die("Can't read p-value of sequence %s.", sequence_id);
}
if (sscanf((c=strtok(NULL, " "))?c:"", "%d", &num_occurs) != 1) {
die("Can't read number of motif occurences in sequence %s.",
sequence_id);
}
if (sscanf((c=strtok(NULL, " "))?c:"", "%d", &seq_length) != 1) {
die("Can't read length of sequence %s.", sequence_id);
}
if (verbosity > NORMAL_VERBOSE) {
fprintf(stderr, "Reading motif occurrences for sequence %s.\n",
sequence_id);
}
// If requested, try to create an order string.
if (find_order) {
new_order = create_empty_order(num_occurs, sequence_p);
}
// Accumulate motif occurence data.
sprintf(prev_motif, "%d", 0);
prev_position = 0;
for (i_occur = 0; i_occur < num_occurs; i_occur++) {
char motif_id[MAX_MOTIF_ID_LENGTH + 1]; // ID of the current motif.
int motif_position; // Position of the current motif occurrence.
char *c; // Dummy to hold return of strtok.
// Read the three values.
if (sscanf((c=strtok(NULL, " "))?c:"", "%s", motif_id) != 1) {
die("Can't read index of occurrence %d in sequence %s.",
i_occur, sequence_id);
}
if (sscanf((c=strtok(NULL, " "))?c:"", "%d", &motif_position) != 1) {
die("Can't read position of occurrence %d in sequence %s.",
i_occur, sequence_id);
}
if (sscanf((c=strtok(NULL, " "))?c:"", "%f", &motif_p) != 1) {
die("Can't read p-value of occurrence %d in sequence %s.",
i_occur, sequence_id);
}
// Only include motifs that have been retained
if (have_motif(motif_id, num_motifs, motifs)) {
// Make sure we have strand information in the ID.
if (has_reverse_strand) {
add_strand(motif_id);
}
// Record this occurrence.
record_occurrence(sequence_id,
motif_id,
p_threshold,
motif_p,
prev_motif,
&prev_position,
motif_position,
*transp_freq,
*spacer_ave,
new_order,
num_motifs,
motifs);
/* Motifs are stored in order of their motif IDs, but they are
indexed from zero rather than one. */
prev_position = motif_position +
(motifs[find_matrix_location(motifs, motif_id, num_motifs) - 1]).length;
}
}
assert(seq_length >= prev_position);
// Record the transition to the end state.
record_occurrence(sequence_id,
END_TRANSITION,
p_threshold,
motif_p,
prev_motif,
&prev_position,
seq_length,
*transp_freq,
*spacer_ave,
new_order,
num_motifs,
motifs);
// Decide whether to keep the new order object.
if (find_order) {
if ((get_num_distinct(new_order) > get_num_distinct(*order_spacing)) ||
(((get_num_distinct(new_order) == get_num_distinct(*order_spacing))
&& (get_pvalue(new_order) < get_pvalue(*order_spacing))))) {
if (verbosity > NORMAL_VERBOSE) {
fprintf(stderr, "Storing order from sequence %s (%g < %g).\n",
sequence_id, get_pvalue(new_order),
get_pvalue(*order_spacing));
print_order_and_spacing(stderr, new_order);
}
free_order(*order_spacing);
*order_spacing = new_order;
} else {
free_order(new_order);
}
}
}
}
/*
* Using the linreg test,
*
* this method returns the lowest scoring subdivision of a set of sequences for a given motif.
* It's not self-contained, as it requires to hook into the global variables results, motifs, seq_ids.
*/
ramen_result_t* ramen_do_linreg_test(int motif_num) {
//Assorted vars
int seq_num;
int j,k;
int motif_index = motif_num * 2; //This is a workaround to the change in the motif datastructure where it now
// goes +MOTIFA -MOTIFA +MOTIFB etc. rather than all + then all - motifs.
//Vars for the regression
double* x;
double* y;
double m = 0;
double b = 0;
double mse = 0;
//Vars for scoring
ramen_result_t* r;
//Allocate memory or set initial values
seq_num = get_num_strings(seq_ids); //number of sequences
r = malloc(sizeof(ramen_result_t)); //allocate space, as a ptr to this will go in the array later
//that's why we don't free it in this loop.
x = malloc(sizeof(double)*seq_num);
y = malloc(sizeof(double)*seq_num);
//Now we need to copy the scores into two double arrays
//Use LOG macro so that log(0) 'works'
for (j=0; j < seq_num; j++) {
if (args.log_fscores == TRUE) {
y[j] = LOG(get_array_item(j, seq_fscores));
} else {
y[j] = get_array_item(j, seq_fscores);
}
if (args.log_pwmscores == TRUE) {
x[j] = LOG(results[motif_num][j]);
} else {
x[j] = results[motif_num][j];
}
}
//Switch x&y if they're to be switched
if (args.linreg_switchxy) {
SWAP(double*, x, y);
}
// TODO: Tidy and/or remove this for production
if(args.linreg_dump_dir > 0) {
FILE *fh;
char* filename;
filename = malloc(sizeof(char)*(strlen(args.linreg_dump_dir) + 50));
sprintf(filename, "%s/%s.tsv", args.linreg_dump_dir, get_motif_id(motif_at(motifs.motifs, motif_index)));
fh = fopen(filename, "w");
fputs("PWM_Score\tFluorescence_Score\n", fh);
for (j=0; j < seq_num; j++) {
fprintf(fh, "%.10e %.10e\n", x[j], y[j]);
}
fclose(fh);
free(filename);
}
/*extern double regress(
int n, / number of points /
double *x, / x values /
double *y, / y values /
double *m, / slope /
double *b / y intercept /
);*/
mse = regress(seq_num, x, y, &m, &b);
if (args.verbose >= 3) {
printf("LinReg MSE of motif %s on %i seqs: %.4g (m: %.4g b: %.4g)\n",
get_motif_id(motif_at(motifs.motifs, motif_index)), seq_num, mse, m, b);
}
//Add to our motif list if lowest MSE
r->motif_id = strdup(get_motif_id(motif_at(motifs.motifs, motif_index)));
r->m = m; //Not p-values, but they'll do when we re-use this structure...
r->b = b;
r->mse = mse;
r->p = -1;
//Do stochastic sampling if required.
if (args.repeats > 0) {
int repeat_wins = 0;
for (j=0;j<args.repeats;j++) {
double repeat_mse = 0;
shuffle(x,seq_num); //Shuffle and break the associations between x and y
repeat_mse = regress(seq_num, x, y, &m, &b);
//fprintf(stderr, "Motif %d Repeat %d RMSE: %g MSE: %g\n",motif_index,j,repeat_mse,mse);
if (repeat_mse <= mse) {
repeat_wins++;
}
}
r->p = repeat_wins*1.0/ args.repeats*1.0;
}
free(x);
free(y);
return r;
}
/*******************************************************************************
* Return count of TRANSFAC species strings.
******************************************************************************/
int get_transfac_num_species(TRANSFAC_MOTIF_T *motif) {
return motif->species_list == NULL
? 0 : get_num_strings(motif->species_list);
}
/****************************************************************************
* Return a list containing the empirical column frequency distributions
* for all alignments in the input.
*
* Each file in the list of filenames is read and the species list is
* determined. The counts of each occurring column are tallied.
* All files with the same species lists get their counts combined.
*
* The returned list contains one distribution per species list that
* occurs in some alignment.
****************************************************************************/
OBJECT_LIST_T* get_alignment_column_freqs_list
(ALPH_T alph,
STRING_LIST_T* filenames,
BOOLEAN_T remove_allgap_seqs)
{
int file_index;
int num_filenames = get_num_strings(filenames);
ARRAY_T* alignment_column_freqs = NULL;
OBJECT_LIST_T* alignment_column_freqs_list
= new_object_list(equal_string_lists,
(void*)copy_string_list,
free_string_list,
free_array);
// Consider each alignment in turn.
for(file_index = 0; file_index < num_filenames; file_index++) {
char* filename = get_nth_string(file_index, filenames);
if (verbosity >= NORMAL_VERBOSE && !(file_index % 1)) {
fprintf(
stderr,
"Computing column freqs: alignment file number %d of %d total files.\n",
file_index+1, num_filenames
);
}
// Read the alignment
int ref_seq_index = 0;
ALIGNMENT_T* alignment =
read_alignment_from_file(filename, TRUE, remove_allgap_seqs, &ref_seq_index);
STRING_LIST_T* alignment_species = get_species_names(alignment);
// Try to retrieve the counts so far for this list of species.
alignment_column_freqs =
(ARRAY_T*)retrieve_object(
alignment_species,
alignment_column_freqs_list
);
// Found counts for current species list?
if (alignment_column_freqs) {
// Add counts from current alignment.
(void) build_alignment_column_counts(alph, alignment, alignment_column_freqs);
// Note: objects in lists are references, so no need to re-store
// after modification.
}
// Didn't find counts for this species list, so create new array of counts.
else {
alignment_column_freqs = build_alignment_column_counts(alph, alignment, NULL);
store_object(
(void*)alignment_column_freqs,
(void*)alignment_species,
0.0, // Score
alignment_column_freqs_list
);
}
// free space used by alignment
free_alignment(alignment);
} // each filename
fprintf(stderr, "\n");
// Convert counts to frequencies by retrieving each array of counts
// and dividing by the total counts for that list of species.
while (
(alignment_column_freqs = retrieve_next_object(alignment_column_freqs_list)
) != NULL )
{
int i;
int num_freqs = get_array_length(alignment_column_freqs);
double total_counts;
// Get total counts.
for (i=total_counts=0; i<num_freqs; i++) {
total_counts += get_array_item(i, alignment_column_freqs);
}
// Get frequencies.
for (i=0; i<num_freqs; i++) {
double f = get_array_item(i, alignment_column_freqs);
set_array_item(i, f/total_counts, alignment_column_freqs);
#ifdef DEBUG
int asize = alph_size(alph, ALPH_SIZE);
int num_leaves = NINT(log(num_freqs)/log(asize));
char* alignment_col = mm_malloc((num_leaves + 1) * sizeof(char));
unhash_alignment_col(
alph,
i, //col_index
alignment_col,
num_leaves
);
printf("%s %g %g\n", alignment_col, f, f/total_counts);
myfree(alignment_col);
#endif
} // get frequencies
} // while more species lists
return(alignment_column_freqs_list);
} // get_alignment_column_freqs_list