IntroductionWe're trying to simplify Pygr's data models for the 0.8 release. Here I'll propose how we might refactor the sequence database model a bit. My main goals: - call the base class SequenceDB. Put all BLAST support in a subclass.
- define a standard, modular API to the actual storage, so no storage code is mixed up in the SequenceDB class. It should be possible to use different storage classes with SequenceDB.
- let users define / supply any sequence reading function they want.
- start using the standard get_bound_subclass() system for handling itemClass, the same as in other parts of Pygr.
What's NewThe main change is modularizing the storage access mechanisms into the the seqInfoDict and _init_subclass() classmethod supplied by the itemClass. Proposed SequenceDB model- itemClass attribute: class to use for each top-level sequence object. I propose that all functions for storage access (building and searching the index) be part of the itemClass, since it represents the storage interface.
- itemSliceClass attribute: class to use for sequence sub-slice objects
- seqInfoDict attribute: dictionary interface to sequence information from the storage mechanism. Returns an object with attributes like length, title and possibly others like offset. This is the official mechanism for getting some information about what's in the database without actually triggering the construction of a sequence object. This interface is needed for things like NLMSA that will need to construct union coordinate systems that unify one or more sequence databases.
- __getitem__(seqID): get the sequence object with this ID
- __len__(): get total number of sequences in this database
- __invert__(): get reverse mapping object (maps sequence obj to its ID).
- __contains__(B): True if the argument is a sequence in this database
- __iter__() etc.: all the standard dictionary iterators
- cacheHint(owner, ivalDict): save a cache hint dict of {id:(start,stop)} associated with owner
- strsliceCache(seq,start,stop): get strslice using cache hints, if any
Sequence model- id attribute: gives the ID of the sequence (primary key within its database)
- db attribute: points to the database object containing this sequence
- path: the complete sequence object containing this sequence interval
- start, stop: coordinates for this interval. See docs for sign convention.
- orientation: 1 if forward strand, -1 if negative strand
- __getitem__(slice): slice this sequence
- __len__(): get this sequence's length
- __str__(): get this sequence interval's sequence string
- __neg__(): get reverse-complement of this sequence interval (ValueError if protein)
- __add__(B): get union with another sequence interval, i.e. A+B covers the interval A.start,B.stop. (ValueError if not in the same parent sequence).
- __contains__(B): True if the argument is a sub-interval of this sequence interval
- __mul__(B): get intersection with another sequence interval, i.e. A*B is the largest interval contained both in A and in B.
- before(): get the entire interval up to this sequence (adjacent on the left)
- after(): get the entire interval after this sequence (adjacent on the right)
- seqtype(): get type (nucleotide or protein) of this sequence
Additional implementor methods- strslice(start, stop): get string corresponding to sequence for interval start,stop. This is the primary interface to the actual storage.
- _init_subclass(seqReader=None): classmethod on the itemClass, that initializes connection to the storage, constructing the index if necessary, and adds a seqInfoDict attribute to the sequence database object. Should accept a seqReader method argument that iterates all the sequences in the input file and returns all the IDs and sequences.
|
Sign in to add a comment
