An Integrated Data Analysis Suite and Programming ... - TOBIAS-lib

2.5. VERSATILE OLIGOMER DETECTION AND READ CLIPPING 41
The conguration depicted in the rst column indicates the type of end overhang that should
not be penalized by the alignment algorithm's scoring method, with the lower line dened as
reference (ref) and the upper line as query (qry). The dierent end alignment modes form a
hierarchy of constraints, with dangling_any a subset of local, dangling_qry and dangling_ref
subsets of dangling_any, and global a common subset dangling_qry and dangling_ref.
The term query will be used in the following to always indicate the sequencing read, and
reference may thus refer to a possibly much shorter oligomer sequence. Alignment of a short
read to a part of a genomic reference sequence constitutes an example of overlap conguration 11,
dened by the keyword pair (dangling_ref;dangling_ref). Detection of sequencing adapters
or 3 ′ bar codes in small RNA sequencing data corresponds to congurations 6 or 7, depending
on whether the read contains the entire or only the partial adapter sequence. This subset of
congurations is dened by the pair (dangling_qry;dangling_any). Detection of bar codes in
5 ′ bar-coded sequences is represented by case 2 (global;dangling_qry).
For Cre/lox-type recombinant mate pair sequencing varied overlap constraints may be appropriate
depending on the desired sensitivity-specicity tradeo. Conservative detection of
linker sequences in valid 454 type mate pairs corresponds to conguration 6. Illumina Cre/lox
mate pair protocols obtain read pairs where the linker sequence is expected towards the end
of either read (conguration 6 or 7). Occasionally one of the enzymatic cuts of the circular
DNA may also occur inside the linker DNA. Such cases are described by conguration
10. The subset of congurations 6, 7 and 10 constitutes a case of mutually dependent end
alignment congurations. Thus, it can not be accounted for by a single keyword pair, but the
two pairs (dangling_qry;dangling_any) and (dangling_ref;dangling_qry) (or alternatively,
(dangling_any;dangling_qry) and (dangling_qry;dangling_ref)).
The SHORE oligo-match utility is capable of for each sequencing read selecting the optimal
alignment or alignments out of multiple pairs of end alignment modes, multiple reference
oligomers and optionally their reverse complemented sequence.
By utilizing a fully customizable 16x16 scoring matrix, the program enables adjustable handling
of ambiguous IUPAC nucleotide codes as well as asymmetric base mismatch penalties with
respect to the direction of the match.
A pair-wise sequence mapping is composed of two pairs of end coordinates as well as the
alignment describing actual base pairings and sequence gaps. To increase specicity of read
clipping and splitting operations, it is desirable to ensure that optimal pair-wise mappings feature
a unique pair of end coordinates, whereas potential alternative alignments can be considered
irrelevant. Our alignment algorithm is capable of either generating an exhaustive list of all
possible alignments, a list featuring a representative of all alignments with a dierent pair of
end coordinates, or just a single representative for each pair-wise mapping, which is optionally
assessed for uniqueness of end coordinates.
The utility provides lters for pair-wise mappings with respect to uniqueness of oligomer
selection and end coordinates. Alignments valid following ltering may be comprehensively
reported and applied to clipping or splitting sequencing reads at either, or at both ends of the
detected oligomer.
2.5.2 Dynamic Programming Alignment and Backtracing Pipeline
Our alignment pipeline proceeds in three subsequent passes. Initially dynamic programming
alignment of the sequencing read is performed to each oligomer and for each pair of provided end
alignment modes. The optimal alignment or alignments are passed to the backtracing module.
Finally, generated traces are ltered and may subsequently be applied to read manipulation.