class Numbers(NumbersI, MappedList):
"""Safe version of Numbers that validates on all list operations.
For each item in data (which must be iterable), tests whether the item
is a number and, if so, adds it to the Numbers.
Note: this means we have to override _all_ the list methods that might
potentially add new data to the list. This makes it much slower than
UnsafeNumbers, but impossible for it to hold invalid data.
"""
Mask = FunctionWrapper(float)
def __init__(self, data=None, Constraint=None, Mask=None):
"""Initializes a new Numbers object.
Usage: nl = Numbers(data)
For each item in data, tries to convert to a float. If successful,
produces new Numbers with data.
Note: this means that a single string of digits will be treated as
a list of digits, _not_ as a single number. This might not be what
you expected.
Also, data must be iterable (so a 1-element list containing a number
is OK, but a single number by itself is not OK).
"""
if data is not None:
data = map(float, data) #fails if any items are not floatable
else:
data = []
MappedList.__init__(self, data, Constraint, Mask)
class Freqs(FreqsI, MappedDict):
"""Holds a frequency distribution, i.e. a set of category -> count pairs.
Class data:
ValueMask: function that transforms values before they are entered.
RequiredKeys: keys that are automatically added with frequency 0 before
frequencies are added.
Performs (expensive) validation on many operations that change the
dictionary. Use UnsafeFreqs if speed is more important than validation.
"""
ValueMask = FunctionWrapper(freqwatcher)
def __init__(self, data=None, Constraint=None, Mask=None, ValueMask=None):
"""Passes on to superclass, but adds required keys if absent.
Parameters (for polymorphism with MappedDict superclass):
data: data to load into self
Constraint: only items that Constraint __contains__ are allowed
Mask: function applied to keys before lookup
ValueMask: function applied to values before addition
"""
super(Freqs, self).__init__(Constraint=Constraint, Mask=Mask, \
ValueMask=ValueMask)
self += data
for key in self.RequiredKeys:
if key not in self:
self[key] = 0.0
class DbRefs(MappedRecord, ConstrainedDict):
"""Holds Database -> [Accessions] mapping.
The accessions for a particular database are always stored as a list.
DbRefs will ultimately contain methods for actually getting the records
from known databases.
"""
ValueMask = FunctionWrapper(_make_list)
DefaultValue = []
def __init__(self, parameters={}, synonyms={}):
"""Initialize the Parameters object.
parameters: a dictionary of Parameter objects keyed by their identifier
synonyms: a dictionary of synonyms. Keys are synonyms, values are
parameter identifiers.
"""
mask = FunctionWrapper(_find_synonym(synonyms))
super(Parameters, self).__init__(data=deepcopy(parameters), Mask=mask)
self.__setitem__ = self.setdefault = self.update =\
self.__delitem__ = self._raiseNotImplemented
class NumberFreqs(NumberFreqsI, MappedDict):
"""Class holding freqs where both keys and values are numbers.
Mean, variance etc. assume that the data are frequencies of other
numbers rather than treating each key as a separate category.
Changes calculation of mean, standard deviation, etc. by assuming that
the keys have weight proportional to their values (i.e. if the key is
5 and the value is 3, it contributes 15 'units' rather than 3 to things
like mean() and normalize()).
Performs (expensive) validation to ensure that keys are floats and
values are non-negative floats.
All keys and values are automatically converted to float.
"""
RequiredKeys = None
Mask = FunctionWrapper(float)
ValueMask = FunctionWrapper(freqwatcher)
def __init__(self, data=None, Constraint=None, Mask=None, ValueMask=None):
"""Passes on to superclass, but adds required keys if absent.
Parameters (for polymorphism with MappedDict superclass):
data: data to load into self
Constraint: only items that Constraint __contains__ are allowed
Mask: function applied to keys before lookup
ValueMask: function applied to values before addition
"""
super(NumberFreqs, self).__init__(Constraint=Constraint, Mask=Mask, \
ValueMask=ValueMask)
self += data
r = self.RequiredKeys
if r:
for key in r:
if key not in self:
self[key] = 0.0
class AminoAcidUsage(InfoFreqs):
"""Stores counts ofthe 20 canonical amino acids."""
Mask = FunctionWrapper(upper)
RequiredKeys = dict.fromkeys(AminoAcids)
def bases(self, genetic_code=SGC, codon_usage=_equal_codons):
"""Predicts most likely set of base frequencies.
Optionally uses a genetic code (default: standard genetic code) and
codon usage (default: unbiased codon usage).
"""
result = self.codons(genetic_code, codon_usage).bases()
result.normalize()
return result
def codons(self, genetic_code=SGC, codon_usage=_equal_codons):
"""Predicts most likely set of codon frequencies.
Optionally uses genetic_code (to figure out which codons belong
with each amino acid), and codon_usage (to get most likely codons for
each amino acid). Defaults are the standard genetic code and unbiased
codon frequencies.
"""
result = {}
normalized = Freqs(self)
normalized.normalize()
for aa, aa_freq in normalized.items():
curr_codons = [
c.upper().replace('T', 'U') for c in genetic_code[aa]
]
if not curr_codons:
continue #code might be missing some amino acids?
curr_codon_freqs = Numbers([codon_usage[c] for c in curr_codons])
curr_codon_freqs.normalize()
for codon, c_freq in zip(curr_codons, curr_codon_freqs):
result[codon] = c_freq * aa_freq
return CodonUsage(result, self.info, genetic_code)
def positionalBases(self, genetic_code=SGC, codon_usage=_equal_codons):
"""Predicts most likely set of positional base frequencies.
Optionally uses a genetic code (default: standard genetic code) and
codon usage (default: unbiased codon usage).
"""
return self.codons(genetic_code, codon_usage).positionalBases()
def aminoAcids(self):
"""Supports bases/positionalBases/aminoAcids/codons interface."""
return self
class CodonUsage(CodonUsageI, InfoFreqs):
"""Stores frequencies of the 64 codons, mapped to RNA.
This class is convenient but inefficient, since it automatically maps any
lookups to the uppercase RNA alphabet internally. Use UnsafeBaseUsage for
speed when necessary.
"""
Mask = FunctionWrapper(key_to_rna)
RequiredKeys = RnaCodons
BaseUsageClass = BaseUsage
def __init__(self, data=None, Info=None, GeneticCode=None, \
Mask=None, ValueMask=None, Constraint=None):
"""Initializes new CodonUsage with Info and frequency data.
Note: Mask, ValueMask and Constraint are ignored, but must be present
to support copy() because of the ConstrainedContainer interface.
"""
#check if we have a sequence: if so, take it 3 bases at a time
#this will (properly) fail on lists of tuples or anything else where
#the items don't behave like strings.
try:
codons = [''.join(data[i:i + 3]) for i in xrange(0, len(data), 3)]
except:
codons = data
super(CodonUsage, self).__init__(codons, Info)
if GeneticCode:
if isinstance(GeneticCode, GenCodeClass):
curr_code = GeneticCode
else:
curr_code = GeneticCodes[GeneticCode]
else:
curr_code = self._default_code
self.__dict__['GeneticCode'] = curr_code
def __getitem__(self, key):
"""Normalizes key and treats T=U."""
key = self.Mask(key)
if len(key) == 2: #pair of bases, e.g. GC for GC content
dup = BaseUsage(self)
dup.normalize()
return sum([dup.get(i, 0) for i in key], 0)
else:
return super(CodonUsage, self).__getitem__(key)
class BaseUsage(BaseUsageI, InfoFreqs):
"""Stores frequencies of the four bases, mapped to RNA.
This class is convenient but inefficient, since it automatically maps any
lookups to the uppercase RNA alphabet internally. Use UnsafeBaseUsage for
speed when necessary.
"""
Mask = FunctionWrapper(key_to_rna)
RequiredKeys = dict.fromkeys(Bases)
def __getitem__(self, key):
"""Normalizes key and treats T=U."""
key = self.Mask(key)
if len(key) == 2: #pair of bases, e.g. GC for GC content
dup = BaseUsage(self)
dup.normalize()
return sum([dup.get(i, 0) for i in key], 0)
else:
return super(BaseUsage, self).__getitem__(key)
class DinucUsage(DinucI, InfoFreqs):
"""Stores frequencies of the 16 dinucleotides, mapped to RNA.
This class is convenient but inefficient, since it automatically maps any
lookups to the uppercase RNA alphabet internally. Use UnsafeBaseUsage for
speed when necessary.
"""
Mask = FunctionWrapper(key_to_rna)
RequiredKeys = RnaDinucs
def __init__(self, data=None, Info=None, Overlapping=True, \
GeneticCode=None, Mask=None, ValueMask=None, Constraint=None):
"""Initializes new CodonUsage with Info and frequency data.
Note: Mask, ValueMask and Constraint are ignored, but must be present
to support copy() because of the ConstrainedContainer interface.
"""
#check if we have a sequence: if so, take it 3 bases at a time
#this will (properly) fail on lists of tuples or anything else where
#the items don't behave like strings.
if Mask is not None:
self.Mask = Mask
if ValueMask is not None:
self.ValueMask = ValueMask
try:
data = self.Mask(data)
if Overlapping == '3-1':
range_ = range(2, len(data) - 1, 3)
elif Overlapping:
range_ = range(0, len(data) - 1)
else:
range_ = range(0, len(data) - 1, 2)
dinucs = [''.join(data[i:i + 2]) for i in range_]
except:
dinucs = data
super(DinucUsage, self).__init__(dinucs, Info)
def __getitem__(self, key):
"""Normalizes key and treats T=U."""
key = self.Mask(key)
return super(DinucUsage, self).__getitem__(key)
class SpansOnly(ConstrainedList):
"""List that converts elements to Spans on addition."""
Mask = FunctionWrapper(Span)
_constraint = ClassChecker(Span)
def __init__(self, motifset, Gap=IUPAC_gap, Missing=IUPAC_missing,\
Gaps=None,
Sequence=None, Ambiguities=None,
label=None, Complements=None, Pairs=None, MWCalculator=None, \
add_lower=False, preserve_existing_moltypes=False, \
make_alphabet_group=False, ModelSeq=None):
"""Returns a new MolType object. Note that the parameters are in flux.
Currently:
motifset: Alphabet or sequence of items in the default
alphabet. Does not include degenerates.
Gap: default gap symbol
Missing: symbol for missing data
Gaps: any other symbols that should be treated as gaps (doesn't have
to include Gap or Missing; they will be silently added)
Sequence: Class for constructing sequences.
Ambiguities: dict of char:tuple, doesn't include gaps (these are
hard-coded as - and ?, and added later.
label: text label, don't know what this is used for. Unnecessary?
Complements: dict of symbol:symbol showing how the non-degenerate
single characters complement each other. Used for constructing
on the fly the complement table, incl. support for mustPair and
canPair.
Pairs: dict in which keys are pairs of symbols that can pair
with each other, values are True (must pair) or False (might
pair). Currently, the meaning of GU pairs as 'weak' is conflated
with the meaning of degenerate symbol pairs (which might pair
with each other but don't necessarily, depending on how the
symbol is resolved). This should be refactored.
MWCalculator: f(seq) -> molecular weight.
add_lower: if True (default: False) adds the lowercase versions of
everything into the alphabet. Slated for deletion.
preserve_existing_moltypes: if True (default: False), does not
set the MolType of the things added in **kwargs to self.
make_alphabet_group: if True, makes an AlphabetGroup relating
the various alphabets to one another.
ModelSeq: sequence type for modeling
Note on "Degenerates" versus "Ambiguities": self.Degenerates contains
_only_ mappings for degenerate symbols, whereas self.Ambiguities
contains mappings for both degenerate and non-degenerate symbols.
Sometimes you want one, sometimes the other, so both are provided.
"""
self.Gap = Gap
self.Missing = Missing
self.Gaps = frozenset([Gap, Missing])
if Gaps:
self.Gaps = self.Gaps.union(frozenset(Gaps))
self.label = label
#set the sequence constructor
if Sequence is None:
Sequence = ''.join #safe default string constructor
elif not preserve_existing_moltypes:
Sequence.MolType = self
self.Sequence = Sequence
#set the ambiguities
ambigs = {self.Missing:tuple(motifset)+(self.Gap,),self.Gap:(self.Gap,)}
if Ambiguities:
ambigs.update(Ambiguities)
for c in motifset:
ambigs[c] = (c,)
self.Ambiguities = ambigs
#set Complements -- must set before we make the alphabet group
self.Complements = Complements or {}
if make_alphabet_group: #note: must use _original_ ambiguities here
self.Alphabets = AlphabetGroup(motifset, Ambiguities, \
MolType=self)
self.Alphabet = self.Alphabets.Base
else:
if isinstance(motifset, Enumeration):
self.Alphabet = motifset
elif max(len(motif) for motif in motifset) == 1:
self.Alphabet = CharAlphabet(motifset, MolType=self)
else:
self.Alphabet = Alphabet(motifset, MolType=self)
#set the other properties
self.Degenerates = Ambiguities and Ambiguities.copy() or {}
self.Degenerates[self.Missing] = ''.join(motifset)+self.Gap
self.Matches = make_matches(motifset, self.Gaps, self.Degenerates)
self.Pairs = Pairs and Pairs.copy() or {}
self.Pairs.update(make_pairs(Pairs, motifset, self.Gaps, \
self.Degenerates))
self.MWCalculator = MWCalculator
#add lowercase characters, if we're doing that
if add_lower:
self._add_lowercase()
#cache various other data that make the calculations faster
self._make_all()
self._make_comp_table()
# a gap can be a true gap char or a degenerate character, typically '?'
# we therefore want to ensure consistent treatment across the definition
# of characters as either gap or degenerate
self.GapString = ''.join(self.Gaps)
strict_gap = "".join(set(self.GapString) - set(self.Degenerates))
self.stripDegenerate = FunctionWrapper(
keep_chars(strict_gap+''.join(self.Alphabet)))
self.stripBad = FunctionWrapper(keep_chars(''.join(self.All)))
to_keep = set(self.Alphabet) ^ set(self.Degenerates) - set(self.Gaps)
self.stripBadAndGaps = FunctionWrapper(keep_chars(''.join(to_keep)))
#make inverse degenerates from degenerates
#ensure that lowercase versions also exist if appropriate
inv_degens = {}
for key, val in self.Degenerates.items():
inv_degens[frozenset(val)] = key.upper()
if add_lower:
inv_degens[frozenset(''.join(val).lower())] = key.lower()
for m in self.Alphabet:
inv_degens[frozenset(m)] = m
if add_lower:
inv_degens[frozenset(''.join(m).lower())] = m.lower()
for m in self.Gaps:
inv_degens[frozenset(m)] = m
self.InverseDegenerates = inv_degens
#set array type for modeling alphabets
try:
self.ArrayType = self.Alphabet.ArrayType
except AttributeError:
self.ArrayType = None
#set modeling sequence
self.ModelSeq = ModelSeq
class kp(InfoFreqs):
Mask = FunctionWrapper(int)
RequiredKeys = dict.fromkeys([1, 2, 3])