def setUp(self): """setUp: setup method for all tests""" self.true = Pairs([(0,40),(1,39),(2,38),(3,37),(10,20),\ (11,19),(12,18),(13,17),(26,33),(27,32)]) self.predicted = Pairs([(0,40),(1,39),(2,38),(3,37),(4,36),\ (5,35),(10,22),(11,20),(14,29),(15,28)]) self.seq = ['>seq1\n','agguugaaggggauccgauccacuccccggcuggucaaccu']
def adjust_pairs_from_mapping(pairs, mapping): """Returns new Pairs object with numbers adjusted according to map pairs: list of tuples or Pairs object mapping: dictionary containing mapping of positions from one state to the other (e.g. ungapped to gapped) For example: {0: 0, 1: 1, 2: 3, 3: 4, 4: 6, 5: 7, 6: 9, 7: 10, 8: 12} When the Pairs object corresponds to an ungapped sequence and you want to insert gaps, use a mapping from ungapped to gapped. When the Pairs object corresponds to a gapped sequence and you want to degap it, use a mapping from gapped to ungapped. """ result = Pairs() for x, y in pairs: if x is None: new_x = None elif x not in mapping: continue else: new_x = mapping[x] if y is None: new_y = None elif y not in mapping: continue else: new_y = mapping[y] result.append((new_x, new_y)) return result
def column_parser(lines): """Parser column format""" record = False result = [] struct = [] seq = '' for line in lines: if line.startswith('; ------'): #structure part beginns record = True continue if line.startswith('; ******'): #structure part ends record = False struct = adjust_base(struct,-1) struct = Pairs(struct).directed()#remove duplicates struct.sort() result.append([seq,struct]) struct = [] seq = '' continue if record: sline = line.split() if sline[4] == '.': #skip not paired seq = ''.join([seq,sline[1]]) continue seq = ''.join([seq,sline[1]]) pair = (int(sline[3]),int(sline[4])) #(alignpos,align_bp) struct.append(pair) return result
def ilm_parser(lines=None,pseudo=True): """Ilm format parser Takes lines as input and returns a list with Pairs object. Pseudo - if True returns pairs with possible pseudoknot if False removes pseudoknots """ pairs = [] for line in lines: if line.startswith('Final') or len(line)==1:#skip these lines continue line = line.strip('\n') line = map(int,line.split(None,2)) if line[1] == 0: continue #Skip this line, not a pair else: pairs.append(line) pairs = adjust_base(pairs,-1) tmp = Pairs(pairs).directed() tmp.sort() if not pseudo: tmp = opt_single_random(tmp) tmp.sort() result = [] result.append(tmp) return result
def adjust_base(pairs, offset): """Returns new Pairs with values shifted by offset pairs: Pairs object or list of tuples offset: integer Adjusts the base of a pairs object or a list of pairs according to the given offset. There's no validation in here! It is possible negative values are returned -> user responsibility. This method treats all pairs as equal. It'll return a pairs object of exactly the same length as the input, including pairs containing None, and duplicates. Example: adjust_base(Pairs([(2,8),(4,None)]), 2) --> [(4,10),(6,None)] """ if not isinstance(offset, int): raise PairsAdjustmentError("adjust_base: offset should be integer") result = Pairs() for x, y in pairs: if x is not None: new_x = x + offset else: new_x = x if y is not None: new_y = y + offset else: new_y = y result.append((new_x, new_y)) assert len(result) == len(pairs) return result
def test_compare_pairs(self): """compare_pairs: should work on simple case""" #all the same p1 = Pairs([(3,10),(4,9),(5,8),(20,24)]) p2 = Pairs([(3,10),(4,9),(5,8),(20,24)]) self.assertEqual(compare_pairs(p1,p2),1) #all different p1 = Pairs([(3,10),(4,9),(5,8),(20,24)]) p2 = Pairs([(1,2),(3,4),(5,6)]) self.assertEqual(compare_pairs(p1,p2),0) #one empty p1 = Pairs([(3,10),(4,9),(5,8),(20,24)]) p2 = Pairs([]) self.assertEqual(compare_pairs(p1,p2),0) #partially different p1 = Pairs([(1,2),(3,4),(5,6),(7,8)]) p2 = Pairs([(1,2),(3,4),(9,10),(11,12)]) self.assertFloatEqual(compare_pairs(p1,p2),.33333333333333333) #partially different p1 = Pairs([(1,2),(3,4),(5,6)]) p2 = Pairs([(1,2),(3,4),(9,10)]) self.assertFloatEqual(compare_pairs(p1,p2),.5)
def adjust_pairs_from_mapping(pairs, mapping): """Returns new Pairs object with numbers adjusted according to map pairs: list of tuples or Pairs object mapping: dictionary containing mapping of positions from one state to the other (e.g. ungapped to gapped) For example: {0: 0, 1: 1, 2: 3, 3: 4, 4: 6, 5: 7, 6: 9, 7: 10, 8: 12} When the Pairs object corresponds to an ungapped sequence and you want to insert gaps, use a mapping from ungapped to gapped. When the Pairs object corresponds to a gapped sequence and you want to degap it, use a mapping from gapped to ungapped. """ result = Pairs() for x,y in pairs: if x is None: new_x = None elif x not in mapping: continue else: new_x = mapping[x] if y is None: new_y = None elif y not in mapping: continue else: new_y = mapping[y] result.append((new_x, new_y)) return result
def test_sensitivity_dupl(self): """sensitivity: should handle duplicates, pseudo, None""" ref = Pairs([(1,6),(2,5),(3,10),(7,None),(None,None),(5,2),(4,9)]) pred = Pairs([(6,1),(10,11),(3,12)]) self.assertFloatEqual(sensitivity(ref, pred), 0.25) pred = Pairs([(6,1),(10,11),(3,12),(20,None),(None,None),(1,6)]) self.assertFloatEqual(sensitivity(ref, pred), 0.25)
def test_adjust_base_None(self): """adjust_base: should keep Nones or duplicates, ignore conflicts""" pairs = Pairs([(2,8),(3,7),(6,None),(None,None),(2,10)]) expected = Pairs([(1,7),(2,6),(5,None),(None, None),(1,9)]) self.assertEqual(adjust_base(pairs,-1), expected) p = Pairs([(1,2),(2,1),(1,2),(2,None)]) self.assertEqual(adjust_base(p, 1), [(2,3),(3,2),(2,3),(3,None)])
def test_symmetric(self): """Pairs symmetric() should add (down,up) for each (up,down)""" self.assertEqual(self.Empty.symmetric(), []) self.assertEqualItems(self.OneTuple.symmetric(), [(2, 1), (1, 2)]) self.assertEqualItems( Pairs([(1, 2), (1, 2)]).symmetric(), [(1, 2), (2, 1)]) self.assertEqualItems(Pairs([(1,2),(3,4)]).symmetric(),\ [(1,2),(2,1),(3,4),(4,3)]) self.assertEqualItems(Pairs([(1, None)]).symmetric(), [])
def test_ungapped_to_gapped(self): """ungapped_to_gapped: Sequence, ModelSequence, old_cogent, string """ p = Pairs([(0, 6), (1, 5), (3, 9)]) exp = Pairs([(0, 5), (1, 4), (3, 7)]) f = ungapped_to_gapped self.assertEqual(f(self.rna1, exp)[1], p) self.assertEqual(f(self.m1, exp)[1], p) self.assertEqual(f(self.s1, exp)[1], p)
def test_delete_gaps_from_pairs_weird(self): """delete_gaps_from_pairs: should ignore conflicts etc""" r = delete_gaps_from_pairs gap_list = [0, 1, 4, 5, 7, 9] p = Pairs([(2, 6), (3, 8)]) self.assertEqualItems(r(p, gap_list), [(0, 2), (1, 3)]) p = Pairs([(2, 6), (3, 8), (3, None), (6, 2), (3, 8), (None, None)]) self.assertEqualItems(r(p, gap_list),\ [(0,2),(1,3),(1,None),(2,0),(1,3),(None, None)])
def test_sensitivity_empty(self): """sensitivity: should work on emtpy Pairs""" # both empty self.assertFloatEqual(sensitivity(Pairs(), Pairs()), 1) pred = Pairs([(6, 1), (10, 11), (3, 12), (13, 20), (14, 19), (15, 18)]) # prediction emtpy self.assertFloatEqual(sensitivity(Pairs(), pred), 0) # reference empty self.assertFloatEqual(sensitivity(pred, Pairs()), 0)
def test_selectivity_general(self): """selectivity: should work in general""" ref = Pairs([(1, 6), (2, 5), (10, 13)]) pred = Pairs([(6, 1), (3, 4), (10, 12)]) # one good prediction self.assertFloatEqual(selectivity(ref, pred), 0.5) # over-prediction not penalized pred = Pairs([(6, 1), (10, 11), (3, 12), (13, 20), (14, 19), (15, 18)]) self.assertFloatEqual(selectivity(ref, pred), 0.25)
def test_selectivity_empty(self): """selectivity: should handle empty reference/predicted structure""" # both empty self.assertFloatEqual(selectivity(Pairs(), Pairs()), 1) pred = Pairs([(6, 1), (10, 11), (3, 12), (13, 20), (14, 19), (15, 18)]) # prediction emtpy self.assertFloatEqual(selectivity(Pairs(), pred), 0) # reference empty self.assertFloatEqual(selectivity(pred, Pairs()), 0)
def test_directed(self): """Pairs directed() should change all pairs so that a<b in (a,b)""" self.assertEqual(self.Empty.directed(), []) res = self.Undirected.directed() res.sort() self.assertEqual(res, Pairs([(1, 2), (1, 7), (3, 8), (4, 6)])) res = self.UndirectedNone.directed() self.assertEqual(res, Pairs([])) res = self.UndirectedDouble.directed() self.assertEqual(res, Pairs([(1, 2)]))
def test_get_counts_pseudo(self): """get_counts: should work when pseudo in ref -> classification off""" # pairs that would normally be compatible, are now contradicting ref = Pairs([(0, 8), (1, 7), (4, 10)]) pred = Pairs([(0, 8), (3, 6), (4, 10)]) seq = 'GACUGUGUCAU' exp = {'TP':2,'TN':13-2-1, 'FN':1,'FP':1,\ 'FP_INCONS':0, 'FP_CONTRA':1, 'FP_COMP':0} self.assertEqual(get_counts(ref, pred, split_fp=True,\ sequences=[seq], min_dist=4), exp)
def pairs_union(one, other): """Returns the intersection of one and other one: list of tuples or Pairs object other: list of tuples or Pairs object one and other should map onto a sequence of the same length. """ pairs1 = frozenset(Pairs(one).directed()) #removes duplicates pairs2 = frozenset(Pairs(other).directed()) return Pairs(pairs1 | pairs2)
def test_all_metrics_pseudo(self): """all_metrics: pseudoknot in ref, check against compare_ct.pm""" ref = Pairs([(0, 8), (1, 7), (4, 10)]) pred = Pairs([(0, 8), (3, 6), (4, 10)]) seq = 'GACUGUGUCAU' exp = {'SENSITIVITY':0.6666667, 'SELECTIVITY':0.6666667,\ 'AC':0.6666667, 'CC':0.57575758, 'MCC':0.57575758} obs = all_metrics(ref, pred, seqs=[seq], min_dist=4) self.assertEqualItems(obs.keys(), exp.keys()) for k in exp: self.assertFloatEqual(obs[k], exp[k])
def get_counts(ref, predicted, split_fp=False, sequences=None, min_dist=4): """Return TP, TN, FPcont, FPconf FPcomp, FN counts""" result = dict.fromkeys(['TP','TN','FN','FP',\ 'FP_INCONS','FP_CONTRA','FP_COMP'],0) ref_set = frozenset(Pairs(ref).directed()) pred_set = frozenset(Pairs(predicted).directed()) ref_dict = dict(ref.symmetric()) pred_dict = dict(predicted.symmetric()) tp_pairs = ref_set.intersection(pred_set) fn_pairs = ref_set.difference(pred_set) fp_pairs = pred_set.difference(ref_set) result['TP'] = len(tp_pairs) result['FN'] = len(fn_pairs) result['FP'] = len(fp_pairs) if split_fp: fp_incons = [] fp_contra = [] fp_comp = [] for x,y in fp_pairs: if x in ref_dict or y in ref_dict: #print "Conflicting: %d - %d"%(x,y) fp_incons.append((x,y)) else: five_prime = x three_prime = y contr_found = False for idx in range(x,y+1): if idx in ref_dict and\ (ref_dict[idx] < five_prime or\ ref_dict[idx] > three_prime): #print "Contradicting: %d - %d"%(x,y) contr_found = True fp_contra.append((x,y)) break if not contr_found: #print "Comatible: %d - %d"%(x,y) fp_comp.append((x,y)) result['FP_INCONS'] = len(fp_incons) result['FP_CONTRA'] = len(fp_contra) result['FP_COMP'] = len(fp_comp) assert result['FP_INCONS'] + result['FP_CONTRA'] + result['FP_COMP'] ==\ result['FP'] if sequences: num_possible_pairs = get_all_pairs(sequences, min_dist) result['TN'] = num_possible_pairs - result['TP'] -\ result['FP_INCONS'] - result['FP_CONTRA'] return result
def test_pairs_intersection_duplicates(self): """pairs_intersection: should work on flipped pairs and duplicates """ p1 = Pairs([(3,10),(4,9),(5,8),(20,24)]) p2 = Pairs([(10,3),(4,9),(5,8),(9,4),(4,9),(23,30)]) self.assertEqualItems(pairs_intersection(p1,p2),[(3,10),(4,9),(5,8)]) # Conflicts, duplicates, None, pseudoknots p1 = Pairs([(3,10),(4,9),(5,8),(20,24),(22,26),(3,2),(9,4),(6,None)]) p2 = Pairs([(1,12),(4,9),(5,8)]) self.assertEqualItems(pairs_intersection(p1,p2),\ [(4,9),(5,8)])
def test_insert_gaps_in_pairs(self): """insert_gaps_in_pairs: should work with normal and conflicts""" p = Pairs([(0, 3), (1, 2), (1, 4), (3, None)]) gaps = [0, 1, 4, 5, 7] self.assertEqual(insert_gaps_in_pairs(p, gaps),\ [(2,8),(3,6),(3,9),(8,None)]) p = Pairs([(0, 6), (1, 5), (2, None), (3, 7), (0, 1), (5, 1)]) gaps = [0, 2, 6, 9] self.assertEqual(insert_gaps_in_pairs(p, gaps),\ [(1,10),(3,8),(4,None),(5,11),(1,3),(8,3)]) gaps = [2, 3, 4, 9] self.assertEqual(insert_gaps_in_pairs(p, gaps),\ [(0,10),(1,8),(5,None),(6,11),(0,1),(8,1)])
def test_compare_random_to_correct(self): """comapre_random_to_correct: should return correct fraction """ p1 = Pairs([(1, 8), (2, 7), (3, 6), (4, 5)]) p2 = Pairs([(1, 8)]) p3 = Pairs([(1, 8), (2, 7), (4, 5)]) p4 = Pairs([(1, 8), (2, 7), (9, 10), (11, 12)]) self.assertFloatEqual(compare_random_to_correct(p2, p1), 1) self.assertFloatEqual(compare_random_to_correct(p3, p1), 1) self.assertFloatEqual(compare_random_to_correct(p4, p1), 0.5) self.assertFloatEqual(compare_random_to_correct([], p1), 0) self.assertFloatEqual(compare_random_to_correct(p2, []), 0) self.assertFloatEqual(compare_random_to_correct([], []), 1)
def test_adjust_pairs_from_mapping_confl(self): """adjust_pairs_from_mapping: should handle conflicts, pseudo, dupl """ f = adjust_pairs_from_mapping p = Pairs([(0,6),(1,5),(2,None),(None,None),(1,4),(3,7),(6,0)]) m = {0:1,1:3,2:6,3:7,4:8,5:10,6:11,7:12} exp = Pairs([(1,11),(3,10),(6,None),(None,None),(3,8),(7,12),(11,1)]) self.assertEqual(f(p, m), exp) p = Pairs([(1,11),(3,10),(7,12),(6,None),(None,None),(5,8)]) m = {1: 0, 3: 1, 6: 2, 7: 3, 8: 4, 10: 5, 11: 6, 12: 7} exp = Pairs([(0,6),(1,5),(3,7),(2,None),(None,None)]) self.assertEqual(f(p,m), exp)
def ct_parser(lines=None): """Ct format parser Takes lines from a ct file as input Returns a list containing sequence,structure and if available the energy. [[seq1,[struct1],energy1],[seq2,[struct2],energy2],...] """ count = 0 length = '' energy = None seq = '' struct = [] result = [] for line in lines: count+=1 sline = line.split(None,6) #sline = split line if count==1 or new_struct(line):#first line or new struct line. if count > 1: struct = adjust_base(struct,-1) struct = Pairs(struct).directed() struct.sort() if energy is not None: result.append([seq,struct,energy]) energy = None else: result.append([seq,pairs]) struct = [] seq = '' #checks if energy for predicted struct is given if sline.__contains__('dG') or sline.__contains__('ENERGY'): energy = atof(sline[3]) if sline.__contains__('Structure'): energy = atof(sline[2]) else: seq = ''.join([seq,sline[1]]) if not int(sline[4]) == 0:#unpaired base pair = ( int(sline[0]),int(sline[4]) ) struct.append(pair) #structs are one(1) based, adjust to zero based struct = adjust_base(struct,-1) struct = Pairs(struct).directed() struct.sort() if energy is not None: result.append([seq,struct,energy]) else: result.append([seq,struct]) return result
def test_delete_gaps_from_pairs(self): """delete_gaps_from_pairs: should work on standard input""" r = delete_gaps_from_pairs # empty list p = Pairs([]) self.assertEqual(r(p, [1, 2, 3]), []) # normal list p1 = Pairs([(2, 8), (3, 6)]) gap_list = [0, 1, 4, 5, 7, 9] self.assertEqualItems(r(p1, gap_list), [(0, 3), (1, 2)]) p2 = Pairs([(2, 8), (3, 6), (4, 9)]) self.assertEqualItems(r(p2, gap_list), [(0, 3), (1, 2)]) p3 = Pairs([(2, 8), (3, 6), (4, 10)]) self.assertEqualItems(r(p3, gap_list), [(0, 3), (1, 2)])
def test_mismatches(self): """Pairs mismatches() should return #pairs that can't be formed""" # with plain string self.assertEqual(Pairs([(0, 1)]).mismatches('AC', {}), 1) self.assertEqual( Pairs([(0, 1)]).mismatches('AC', {('A', 'C'): None}), 0) self.assertEqual( Pairs([(0, 1)]).mismatches('AC', {('A', 'G'): None}), 1) self.assertEqual(Pairs([(0,1),(2,3),(3,1)]).\ mismatches('ACGU',{('A','U'):None}),3) # using sequence with alphabet sequence = Rna('ACGUA') self.assertEqual( Pairs([(0, 1), (0, 4), (0, 3)]).mismatches(sequence), 2)
def test_pairs_intersection(self): """pairs_intersection: should work on simple case """ p1 = Pairs([(3, 10), (4, 9), (5, 8), (20, 24)]) p2 = Pairs([(1, 12), (4, 9), (5, 8)]) self.assertEqualItems(pairs_intersection(p1, p2), [(4, 9), (5, 8)]) #works when one is empty p1 = Pairs([(3, 10), (4, 9), (5, 8), (20, 24)]) p2 = Pairs([]) self.assertEqualItems(pairs_intersection(p1, p2), []) #works also on lists (not Pairs) p1 = [(3, 10), (4, 9), (5, 8), (20, 24)] p2 = [(1, 12), (4, 9), (5, 8)] self.assertEqualItems(pairs_intersection(p1, p2), [(4, 9), (5, 8)])
def compare_random_to_correct(one, other): """Returns fraction of bp in one that is in other (correct) one: list of tuples or Pairs object other: list of tuples or Pairs object Note: the second structure is the one compared against (the correct structure) """ if not one and not other: return 1.0 if not one or not other: return 0.0 pairs1 = frozenset(Pairs(one).directed()) #removes duplicates pairs2 = frozenset(Pairs(other).directed()) return len(pairs1 & pairs2) / len(pairs1)
def test_adjust_base(self): """adjust_base: should work for pairs object or list of pairs""" p = Pairs() self.assertEqual(adjust_base(p, 10), []) pairs = [(1, 21), (2, 15), (3, 13), (4, 11), (5, 10), (6, 9)] offset = -1 expected = [(0, 20), (1, 14), (2, 12), (3, 10), (4, 9), (5, 8)] obs_pairs = adjust_base(pairs, offset) self.assertEqual(obs_pairs, expected) pairs = Pairs([(0, 10), (1, 9)]) self.assertEqual(adjust_base(pairs, -1), Pairs([(-1, 9), (0, 8)])) self.assertEqual(adjust_base(pairs, 5), Pairs([(5, 15), (6, 14)])) self.assertRaises(PairsAdjustmentError, adjust_base, pairs, 3.5)
def common(structs): """ Will return a list of sequences and structures with the most common structure first in the list. (rest or list unordered!) Don't care which of the sequences for the "winning" sequence that is reported since the are not ranked amongst them self. """ frequency = {} v = 0 indx = 0 result = [] tmp_list = [ ] #lookup the seq for the structures,dont care which winner seq key = [] for block in structs: tmp_list.extend(block) p = tuple(block[-1]) if frequency.__contains__( p): #everytime struct p appears count up by 1 frequency[p] += 1 else: frequency[p] = 1 nr = frequency[p] if nr > v: #Which struct appears most times v = nr key = p #if winning structure has frequency == 1 all structure apper only once if frequency[key] == 1: longest = 0 for block in structs: l = len(block[-1]) if l > longest: #pick longest sequence as the winner key = tuple(block[-1]) winner = Pairs(key) indx = tmp_list.index(winner) - 1 result.append([tmp_list[indx], winner]) #adds the most common structure first del frequency[key] for i in frequency.keys(): #rest of structures added i = Pairs(i) indx = tmp_list.index(i) - 1 result.append([tmp_list[indx], i]) return result
def parse_residues(residue_lines, num_base, unpaired_symbol): """Return RnaSequence and Pairs object from residue lines. residue_lines -- list of lines or anything that behaves like it. Lines should contain: residue_position, residue_identiy, residue_partner. num_base -- int, basis of the residue numbering. In bpseq files from the CRW website, the numbering starts at 1. unpaired_symbol -- string, symbol in the 'partner' column that indicates that a base is unpaired. In bpseq files from the CRW website, the unpaired_symbol is '0'. This parameter should be a string to allow other symbols that can't be casted to an integer to indicate unpaired bases. Checks for double entries both in the sequence and the structure, and checks that the structre is valid in the sense that if (up,down) in there, that (down,up) is the same. """ #create dictionary/list for sequence and structure seq_dict = {} pairs = Pairs() for line in residue_lines: try: pos, res, partner = line.strip().split() if partner == unpaired_symbol: # adjust pos, not partner pos = int(pos) - num_base partner = None else: # adjust pos and partner pos = int(pos) - num_base partner = int(partner) - num_base pairs.append((pos, partner)) #fill seq_dict if pos in seq_dict: raise BpseqParseError(\ "Double entry for residue %s (%s in bpseq file)"\ %(str(pos), str(pos+1))) else: seq_dict[pos] = res except ValueError: raise BpseqParseError("Failed to parse line: %s" % (line)) #check for conflicts, remove unpaired bases if pairs.hasConflicts(): raise BpseqParseError("Conflicts in the list of basepairs") pairs = pairs.directed() pairs.sort() # construct sequence from seq_dict seq = RnaSequence(construct_sequence(seq_dict)) return seq, pairs
def test_tuples(self): """Pairs tuples() should transform the elements of list to tuples""" x = Pairs([]) x.tuples() assert x == [] x = Pairs([[1,2],[3,4]]) x.tuples() assert x == [(1,2),(3,4)] x = Pairs([(1,2),(3,4)]) x.tuples() assert x == [(1,2),(3,4)] assert x != [[1,2],[3,4]]
def compare_pairs(one, other): """Returns size of intersection divided by size of union between two Pairs Use as a similiraty measure for comparing secondary structures. Returns the number of base pairs common to both structures divided by the number of base pairs that is in one or the other structure: (A AND B)/(A OR B) (intersection/union) one: list of tuples or Pairs object other: list of tuples or Pairs object """ if one.hasConflicts() or other.hasConflicts(): raise ValueError("Can't handle conflicts in the structure" "") if not one and not other: return 1.0 pairs1 = frozenset(Pairs(one).directed()) #removes duplicates pairs2 = frozenset(Pairs(other).directed()) return len(pairs1 & pairs2) / len(pairs1 | pairs2)
def compare_pairs_mapping(one, other, one_to_other): """Returns intersection/union given a mapping from the first pairs to second Use in case the numbering of the two Pairs object don't correspond. Sort of aligning two ungapped sequences and comparing their Pairs object via a mapping. one: list of tuples or Pairs object other: list of tuples or Pairs object one_to_other: mapping of positions in first pairs object to positions in second pairs object. For example: # pos in first seq, base, pos in second seq #1 U 0 #2 C 1 #3 G 2 #4 A 3 # A 4 #5 C 5 #6 C 6 #7 U #8 G 7 mapping = {1:0, 2:1, 3:2, 4:3, 5:5, 6:6, 7:None, 8:7} """ if not one and not other: return 1.0 just_in_first = 0 just_in_second = 0 in_both = 0 pairs1 = Pairs(one).directed() #removes duplicates pairs2 = Pairs(other).directed() for x,y in pairs1: other_match = (one_to_other[x],one_to_other[y]) if other_match in pairs2: in_both += 1 pairs2.remove(other_match) else: just_in_first += 1 just_in_second += len(pairs2) return in_both/(just_in_first + in_both + just_in_second)
def parse_residues(residue_lines, num_base, unpaired_symbol): """Return RnaSequence and Pairs object from residue lines. residue_lines -- list of lines or anything that behaves like it. Lines should contain: residue_position, residue_identiy, residue_partner. num_base -- int, basis of the residue numbering. In bpseq files from the CRW website, the numbering starts at 1. unpaired_symbol -- string, symbol in the 'partner' column that indicates that a base is unpaired. In bpseq files from the CRW website, the unpaired_symbol is '0'. This parameter should be a string to allow other symbols that can't be casted to an integer to indicate unpaired bases. Checks for double entries both in the sequence and the structure, and checks that the structre is valid in the sense that if (up,down) in there, that (down,up) is the same. """ #create dictionary/list for sequence and structure seq_dict = {} pairs = Pairs() for line in residue_lines: try: pos, res, partner = line.strip().split() if partner == unpaired_symbol: # adjust pos, not partner pos = int(pos) - num_base partner = None else: # adjust pos and partner pos = int(pos) - num_base partner = int(partner) - num_base pairs.append((pos,partner)) #fill seq_dict if pos in seq_dict: raise BpseqParseError(\ "Double entry for residue %s (%s in bpseq file)"\ %(str(pos), str(pos+1))) else: seq_dict[pos] = res except ValueError: raise BpseqParseError("Failed to parse line: %s"%(line)) #check for conflicts, remove unpaired bases if pairs.hasConflicts(): raise BpseqParseError("Conflicts in the list of basepairs") pairs = pairs.directed() pairs.sort() # construct sequence from seq_dict seq = RnaSequence(construct_sequence(seq_dict)) return seq, pairs
def test_toPartners(self): """Pairs toPartners() should return a Partners object""" a = Pairs([(1,5),(3,4),(6,9),(7,8)]) #normal b = Pairs([(0,4),(2,6)]) #pseudoknot c = Pairs([(1,6),(3,6),(4,5)]) #conflict self.assertEqual(a.toPartners(10),[None,5,None,4,3,1,9,8,7,6]) self.assertEqual(a.toPartners(13,3),\ [None,None,None,None,8,None,7,6,4,12,11,10,9]) assert isinstance(a.toPartners(10),Partners) self.assertEqual(b.toPartners(7),[4,None,6,None,0,None,2]) self.assertRaises(ValueError,c.toPartners,7) self.assertEqual(c.toPartners(7,strict=False),[None,None,None,6,5,4,3]) #raises an error when try to insert something at non-existing indices self.assertRaises(IndexError,c.toPartners,0)
def delete_gaps_from_pairs(pairs, gap_list): """Returns Pairs object with pairs adjusted to gap_list pairs: list of tuples or Pairs object gap_list: list or array of gapped positions that should be removed from the pairs object Base pairs of which one of the partners or both of them are in the gap list are removed. If both of them are not in the gap_list, the numbering is adjusted according to the gap_list. When at least one of the two pair members is in the gap_list, the pair will be removed. The rest of the structure will be left intact. Pairs containing None, duplicates, pseudoknots, and conflicts will be maintained and adjusted according to the gap_list. """ if not gap_list: result = Pairs() result.extend(pairs) return result g = array(gap_list) result = Pairs() for up, down in pairs: if up in g or down in g: continue else: if up is not None: new_up = up - g.searchsorted(up) else: new_up = up if down is not None: new_down = down - g.searchsorted(down) else: new_down = down result.append((new_up, new_down)) return result
def insert_gaps_in_pairs(pairs, gap_list): """Adjusts numbering in pairs according to the gap list. pairs: Pairs object gap_list: list of integers, gap positions in a sequence The main assumptionis that all positions in pairs correspond to ungapped positions. If this is not true, the result will be meaningless. """ if not gap_list: new = Pairs() new.extend(pairs) return new ungapped = [] for idx in range(max(gap_list)+2): if idx not in gap_list: ungapped.append(idx) new = Pairs() for x,y in pairs: if x is not None: try: new_x = ungapped[x] except IndexError: new_x = ungapped[-1] + (x-len(ungapped)+1) else: new_x = x if y is not None: try: new_y = ungapped[y] except IndexError: new_y = ungapped[-1] + (y-len(ungapped)+1) else: new_y = y new.append((new_x, new_y)) return new
class PairsTests(TestCase): """Tests for Pairs object""" def setUp(self): """Pairs SetUp method for all tests""" self.Empty = Pairs([]) self.OneList = Pairs([[1,2]]) self.OneTuple = Pairs([(1,2)]) self.MoreLists = Pairs([[2,4],[3,9],[6,36],[7,49]]) self.MoreTuples = Pairs([(2,4),(3,9),(6,36),(7,49)]) self.MulNoOverlap = Pairs([(1,10),(2,9),(3,7),(4,12)]) self.MulOverlap = Pairs([(1,2),(2,3)]) self.Doubles = Pairs([[1,2],[1,2],[2,3],[1,3]]) self.Undirected = Pairs([(2,1),(6,4),(1,7),(8,3)]) self.UndirectedNone = Pairs([(5,None),(None,3)]) self.UndirectedDouble = Pairs([(2,1),(1,2)]) self.NoPseudo = Pairs([(1,20),(2,19),(3,7),(4,6),(10,15),(11,14)]) self.NoPseudo2 = Pairs([(1,3),(4,6)]) #((.(.)).) self.p0 = Pairs([(0,6),(1,5),(3,8)]) #(.((..(.).).)) self.p1 = Pairs([(0,9),(2,12),(3,10),(5,7)]) #((.(.(.).)).) self.p2 = Pairs([(0,10),(1,9),(3,12),(5,7)]) #((.((.(.)).).)) self.p3 = Pairs([(0,9),(1,8),(3,14),(4,13),(6,11)]) #(.(((.((.))).)).(((.((((..))).)))).) self.p4 = Pairs([(0,35),(2,11),(3,10),(4,9),(6,14),(7,13),(16,28),\ (17,27),(18,26),(20,33),(21,32),(22,31),(23,30)]) #(.((.).)) self.p5 = Pairs([(0,5),(2,8),(3,7)]) self.p6 = Pairs([(0,19),(2,6),(3,5),(8,14),(9,13),(10,12),\ (16,22),(17,21)]) self.p7 = Pairs([(0,20),(2,6),(3,5),(8,14),(9,10),(11,16),(12,15),\ (17,23),(18,22)]) def test_init(self): """Pairs should initalize with both lists and tuples""" self.assertEqual(self.Empty,[]) self.assertEqual(self.OneList,[[1,2]]) self.assertEqual(self.OneTuple,[(1,2)]) self.assertEqual(self.MulNoOverlap,[(1,10),(2,9),(3,7),(4,12)]) self.assertEqual(self.MulOverlap,[(1,2),(2,3)]) def test_toPartners(self): """Pairs toPartners() should return a Partners object""" a = Pairs([(1,5),(3,4),(6,9),(7,8)]) #normal b = Pairs([(0,4),(2,6)]) #pseudoknot c = Pairs([(1,6),(3,6),(4,5)]) #conflict self.assertEqual(a.toPartners(10),[None,5,None,4,3,1,9,8,7,6]) self.assertEqual(a.toPartners(13,3),\ [None,None,None,None,8,None,7,6,4,12,11,10,9]) assert isinstance(a.toPartners(10),Partners) self.assertEqual(b.toPartners(7),[4,None,6,None,0,None,2]) self.assertRaises(ValueError,c.toPartners,7) self.assertEqual(c.toPartners(7,strict=False),[None,None,None,6,5,4,3]) #raises an error when try to insert something at non-existing indices self.assertRaises(IndexError,c.toPartners,0) def test_toVienna(self): """Pairs toVienna() should return a ViennaStructure if possible""" a = Pairs([(1,5),(3,4),(6,9),(7,8)]) #normal b = Pairs([(0,4),(2,6)]) #pseudoknot c = Pairs([(1,6),(3,6),(4,5)]) #conflict d = Pairs([(1,6),(3,None)]) e = Pairs([(1,9),(8,2),(7,3)]) #not directed f = Pairs([(1,6),(2,5),(10,15),(14,11)]) # not directed self.assertEqual(a.toVienna(10),'.(.())(())') self.assertEqual(a.toVienna(13,offset=3),'....(.())(())') self.assertRaises(PairError,b.toVienna,7) #pseudoknot NOT accepted self.assertRaises(Exception,b.toVienna,7) #old test for exception self.assertRaises(ValueError,c.toVienna,7) #pairs containging None are being skipped self.assertEquals(d.toVienna(7),'.(....)') #raises error when trying to insert at non-existing indices self.assertRaises(IndexError,a.toVienna,3) self.assertEqual(Pairs().toVienna(3),'...') #test when parsing in the sequence self.assertEqual(a.toVienna('ACGUAGCUAG'),'.(.())(())') self.assertEqual(a.toVienna(Rna('AACCGGUUAGCUA'), offset=3),\ '....(.())(())') self.assertEqual(e.toVienna(10),'.(((...)))') self.assertEqual(f.toVienna(20),'.((..))...((..))....') def test_tuples(self): """Pairs tuples() should transform the elements of list to tuples""" x = Pairs([]) x.tuples() assert x == [] x = Pairs([[1,2],[3,4]]) x.tuples() assert x == [(1,2),(3,4)] x = Pairs([(1,2),(3,4)]) x.tuples() assert x == [(1,2),(3,4)] assert x != [[1,2],[3,4]] def test_unique(self): """Pairs unique() should remove double occurences of certain tuples""" self.assertEqual(self.Empty.unique(),[]) self.assertEqual(self.MoreTuples.unique(),self.MoreTuples) self.assertEqual(self.Doubles.unique(),Pairs([(1,2),(2,3),(1,3)])) def test_directed(self): """Pairs directed() should change all pairs so that a<b in (a,b)""" self.assertEqual(self.Empty.directed(),[]) res = self.Undirected.directed() res.sort() self.assertEqual(res,Pairs([(1,2),(1,7),(3,8),(4,6)])) res = self.UndirectedNone.directed() self.assertEqual(res,Pairs([])) res = self.UndirectedDouble.directed() self.assertEqual(res,Pairs([(1,2)])) def test_symmetric(self): """Pairs symmetric() should add (down,up) for each (up,down)""" self.assertEqual(self.Empty.symmetric(),[]) self.assertEqualItems(self.OneTuple.symmetric(),[(2,1),(1,2)]) self.assertEqualItems(Pairs([(1,2),(1,2)]).symmetric(),[(1,2),(2,1)]) self.assertEqualItems(Pairs([(1,2),(3,4)]).symmetric(),\ [(1,2),(2,1),(3,4),(4,3)]) self.assertEqualItems(Pairs([(1,None)]).symmetric(),[]) def test_paired(self): """Pairs paired() should omit all pairs containing None""" self.assertEqual(self.Empty.paired(),[]) self.assertEqual(Pairs([(1,2),(2,None),(None,3),(None,None)]).paired()\ ,[(1,2)]) def test_hasConflicts(self): """Pairs hasConflicts() should return True if there are conflicts""" assert not self.Empty.hasConflicts() assert not Pairs([(1,2),(3,4)]).hasConflicts() assert Pairs([(1,2),(2,3)]).hasConflicts() assert Pairs([(1,2),(2,None)]).hasConflicts() def test_mismatches(self): """Pairs mismatches() should return #pairs that can't be formed""" # with plain string self.assertEqual(Pairs([(0,1)]).mismatches('AC',{}),1) self.assertEqual(Pairs([(0,1)]).mismatches('AC',{('A','C'):None}),0) self.assertEqual(Pairs([(0,1)]).mismatches('AC',{('A','G'):None}),1) self.assertEqual(Pairs([(0,1),(2,3),(3,1)]).\ mismatches('ACGU',{('A','U'):None}),3) # using sequence with alphabet sequence = Rna('ACGUA') self.assertEqual(Pairs([(0,1),(0,4),(0,3)]).mismatches(sequence),2) def test_hasPseudoknots(self): """Pairs hasPseudoknots() should return True if there's a pseudoknot""" assert not self.NoPseudo.hasPseudoknots() assert not self.NoPseudo2.hasPseudoknots() #add tests for ((.))() etc assert self.p0.hasPseudoknots() assert self.p1.hasPseudoknots() assert self.p2.hasPseudoknots() assert self.p3.hasPseudoknots() assert self.p4.hasPseudoknots() assert self.p5.hasPseudoknots() assert self.p6.hasPseudoknots() assert self.p7.hasPseudoknots()
def setUp(self): """Pairs SetUp method for all tests""" self.Empty = Pairs([]) self.OneList = Pairs([[1,2]]) self.OneTuple = Pairs([(1,2)]) self.MoreLists = Pairs([[2,4],[3,9],[6,36],[7,49]]) self.MoreTuples = Pairs([(2,4),(3,9),(6,36),(7,49)]) self.MulNoOverlap = Pairs([(1,10),(2,9),(3,7),(4,12)]) self.MulOverlap = Pairs([(1,2),(2,3)]) self.Doubles = Pairs([[1,2],[1,2],[2,3],[1,3]]) self.Undirected = Pairs([(2,1),(6,4),(1,7),(8,3)]) self.UndirectedNone = Pairs([(5,None),(None,3)]) self.UndirectedDouble = Pairs([(2,1),(1,2)]) self.NoPseudo = Pairs([(1,20),(2,19),(3,7),(4,6),(10,15),(11,14)]) self.NoPseudo2 = Pairs([(1,3),(4,6)]) #((.(.)).) self.p0 = Pairs([(0,6),(1,5),(3,8)]) #(.((..(.).).)) self.p1 = Pairs([(0,9),(2,12),(3,10),(5,7)]) #((.(.(.).)).) self.p2 = Pairs([(0,10),(1,9),(3,12),(5,7)]) #((.((.(.)).).)) self.p3 = Pairs([(0,9),(1,8),(3,14),(4,13),(6,11)]) #(.(((.((.))).)).(((.((((..))).)))).) self.p4 = Pairs([(0,35),(2,11),(3,10),(4,9),(6,14),(7,13),(16,28),\ (17,27),(18,26),(20,33),(21,32),(22,31),(23,30)]) #(.((.).)) self.p5 = Pairs([(0,5),(2,8),(3,7)]) self.p6 = Pairs([(0,19),(2,6),(3,5),(8,14),(9,13),(10,12),\ (16,22),(17,21)]) self.p7 = Pairs([(0,20),(2,6),(3,5),(8,14),(9,10),(11,16),(12,15),\ (17,23),(18,22)])
def test_toVienna(self): """Pairs toVienna() should return a ViennaStructure if possible""" a = Pairs([(1,5),(3,4),(6,9),(7,8)]) #normal b = Pairs([(0,4),(2,6)]) #pseudoknot c = Pairs([(1,6),(3,6),(4,5)]) #conflict d = Pairs([(1,6),(3,None)]) e = Pairs([(1,9),(8,2),(7,3)]) #not directed f = Pairs([(1,6),(2,5),(10,15),(14,11)]) # not directed self.assertEqual(a.toVienna(10),'.(.())(())') self.assertEqual(a.toVienna(13,offset=3),'....(.())(())') self.assertRaises(PairError,b.toVienna,7) #pseudoknot NOT accepted self.assertRaises(Exception,b.toVienna,7) #old test for exception self.assertRaises(ValueError,c.toVienna,7) #pairs containging None are being skipped self.assertEquals(d.toVienna(7),'.(....)') #raises error when trying to insert at non-existing indices self.assertRaises(IndexError,a.toVienna,3) self.assertEqual(Pairs().toVienna(3),'...') #test when parsing in the sequence self.assertEqual(a.toVienna('ACGUAGCUAG'),'.(.())(())') self.assertEqual(a.toVienna(Rna('AACCGGUUAGCUA'), offset=3),\ '....(.())(())') self.assertEqual(e.toVienna(10),'.(((...)))') self.assertEqual(f.toVienna(20),'.((..))...((..))....')