"""Generic functionality useful for all gene representations. This module contains classes which can be used for all the different types of patterns available for representing gene information (ie. motifs, signatures and schemas). These are the general classes which should be handle any of the different specific patterns. """ # standard library import random # biopython from Bio import utils from Bio.Seq import Seq, MutableSeq class PatternIO: """Allow reading and writing of patterns to files. This just defines a simple persistance class for patterns, making it easy to write them to a file and read 'em back. """ def __init__(self, alphabet = None): """Intialize the reader and writer class. Arguments: o alphabet - An optional argument specifying the alphabet which patterns should follow. If an alphabet is set it'll be used to verify that all patterns follow it. Attributes: o separator - A character to use in separating items in a signature when it is written to a file and read back. This character should not be in the possible alphabet of the sequences, or there will be trouble. """ self._alphabet = alphabet self.separator = ";" def write(self, pattern_list, output_handle): """Write a list of patterns to the given handle. """ for pattern in pattern_list: # deal with signatures, concatentate them with the separator if (type(pattern) == type([]) or type(pattern) == type(tuple([]))): string_pattern = self.separator.join(pattern) # deal with the normal cases else: string_pattern = pattern output_handle.write("%s\n" % string_pattern) def write_seq(self, seq_pattern_list, output_handle): """Convenience function to write Seq objects to a file. This can take Seqs and MutableSeqs, and write them to a file as strings. """ # convert the seq patterns into just string patterns all_patterns = [] for seq_pattern in seq_pattern_list: if isinstance(seq_pattern, MutableSeq): seq = seq_pattern.toseq() all_patterns.append(seq.tostring()) elif isinstance(seq_pattern, Seq): all_patterns.append(seq_pattern.tostring()) else: raise ValueError("Unexpected pattern type %r" % seq_pattern) self.write(all_patterns, output_handle) def read(self, input_handle): """Read patterns from the specified handle. """ all_patterns = [] while 1: cur_line = input_handle.readline() if not(cur_line): break cur_pattern = cur_line.rstrip() # split up signatures if cur_pattern.find(self.separator) >= 0: cur_pattern = tuple(cur_pattern.split(self.separator)) if self._alphabet is not None: # make single patterns (not signatures) into lists, so we # can check signatures and single patterns the same if type(cur_pattern) != type(tuple([])): test_pattern = [cur_pattern] else: test_pattern = cur_pattern for pattern_item in test_pattern: pattern_seq = Seq(pattern_item, self._alphabet) if not(utils.verify_alphabet(pattern_seq)): raise ValueError("Pattern %s not matching alphabet %s" % (cur_pattern, self._alphabet)) all_patterns.append(cur_pattern) return all_patterns class PatternRepository: """This holds a list of specific patterns found in sequences. This is designed to be a general holder for a set of patterns and should be subclassed for specific implementations (ie. holding Motifs or Signatures. """ def __init__(self, pattern_info): """Initialize a repository with patterns, Arguments: o pattern_info - A representation of all of the patterns found in a *Finder search. This should be a dictionary, where the keys are patterns, and the values are the number of times a pattern is found. The patterns are represented interally as a list of two tuples, where the first element is the number of times a pattern occurs, and the second is the pattern itself. This makes it easy to sort the list and return the top N patterns. """ self._pattern_dict = pattern_info # create the list representation self._pattern_list = [] for pattern_name in self._pattern_dict: self._pattern_list.append((self._pattern_dict[pattern_name], pattern_name)) self._pattern_list.sort() self._pattern_list.reverse() def get_all(self): """Retrieve all of the patterns in the repository. """ patterns = [] for pattern_info in self._pattern_list: patterns.append(pattern_info[1]) return patterns def get_random(self, num_patterns): """Retrieve the specified number of patterns randomly. Randomly selects patterns from the list and returns them. Arguments: o num_patterns - The total number of patterns to return. """ all_patterns = [] while len(all_patterns) < num_patterns: # pick a pattern, and only add it if it is not already present new_pattern_info = random.choice(self._pattern_list) if new_pattern_info[1] not in all_patterns: all_patterns.append(new_pattern_info[1]) return all_patterns def get_top_percentage(self, percent): """Return a percentage of the patterns. This returns the top 'percent' percentage of the patterns in the repository. """ all_patterns = self.get_all() num_to_return = int(len(all_patterns) * percent) return all_patterns[:num_to_return] def get_top(self, num_patterns): """Return the specified number of most frequently occurring patterns Arguments: o num_patterns - The number of patterns to return. """ all_patterns = [] for pattern_info in self._pattern_list[:num_patterns]: all_patterns.append(pattern_info[1]) return all_patterns def get_differing(self, top_num, bottom_num): """Retrieve patterns that are at the extreme ranges. This returns both patterns at the top of the list (ie. the same as returned by get_top) and at the bottom of the list. This is especially useful for patterns that are the differences between two sets of patterns. Arguments: o top_num - The number of patterns to take from the top of the list. o bottom_num - The number of patterns to take from the bottom of the list. """ all_patterns = [] # first get from the top of the list for pattern_info in self._pattern_list[:top_num]: all_patterns.append(pattern_info[1]) # then from the bottom for pattern_info in self._pattern_list[-bottom_num:]: all_patterns.append(pattern_info[1]) return all_patterns def remove_polyA(self, at_percentage = .9): """Remove patterns which are likely due to polyA tails from the lists. This is just a helper function to remove pattenrs which are likely just due to polyA tails, and thus are not really great motifs. This will also get rid of stuff like ATATAT, which might be a useful motif, so use at your own discretion. XXX Could we write a more general function, based on info content or something like that? Arguments: o at_percentage - The percentage of A and T residues in a pattern that qualifies it for being removed. """ remove_list = [] # find all of the really AT rich patterns for pattern_info in self._pattern_list: pattern_at = float(pattern_info[1].count('A') + pattern_info[1].count('T')) / len(pattern_info[1]) if pattern_at > at_percentage: remove_list.append(pattern_info) # now remove them from the master list for to_remove in remove_list: self._pattern_list.remove(to_remove) def count(self, pattern): """Return the number of times the specified pattern is found. """ try: return self._pattern_dict[pattern] except KeyError: return 0