View Single Post
Old 12-08-2011, 02:16 PM   #4
gringer
David Eccles (gringer)
 
Location: Wellington, New Zealand

Join Date: May 2011
Posts: 838
Default

Sequencer reads have a chance of read error (e.g. spot misidentification), combined with a chance of sequence error (e.g. polymerase misread in the PCR step). For sequencers that output in base space, both these errors have a similar effect on the base-space mapping. For sequencers that output in color-space, the read errors result in a somewhat unexpected base-space translation even if the underlying sequence has a perfect match to the reference. The issues relating to color-space to base-space translation were discussed in the thread you linked to, but here's my take on it (dumped from an email I recently sent to someone else):

A color-space sequence is an encoding of adjacent dimers such that unchanging bases are encoded with '0', complementary changes are encoded with '3', the colour '1' is used for a non-complementary base change on the same side of the alphabet (AC, CA, GT, or TG), and the colour '2' is used for a non-complementary base change on a different side of the alphabet (AG, GA, CT, or TC). A table of these changes can be found here:

http://www.ploscompbiol.org/article/...i.1000386.g002

This has a few nice properties (e.g. the reverse-complement of a color-space sequence is the same as the reverse of the color-space sequence, a SNP will have two transitions), but many annoying and nasty properties. The first is that a color-space sequence in itself is meaningless without a base reference (usually the starting base).

Here's an example color-space sequence:

Code:
2112322311010133121320003202203201302321 [40 chars]
That color-space sequence can describe four different base-space sequences:

Code:
0: AGTGATCTACAACCATACTGCTTTTAGGAGGCTTGCCTAGT [41 chars]
1: CTGTCGAGCACCAACGCAGTAGGGGCTTCTTAGGTAAGCTG
2: GACAGCTCGTGGTTGCGTCATCCCCGAAGAATCCATTCGAC
3: TCACTAGATGTTGGTATGACGAAAATCCTCCGAACGGATCA
Note that these sequences are one character longer than their color-space equivalent, so by adding a starting base the sequence length does not change from the base-space representation.

Code:
A2112322311010133121320003202203201302321 [41 chars, decoded sequence 0]
Here's another annoying property. It was pointed out before that the reverse complement of a color-space sequence is the reverse of the sequence. This is not entirely true if you include the starting base, because that base has now shifted to the end, and is its complementary partner:
Code:
rc(A2112322311010133121320003202203201302321)
=1232031023022023000231213310101132232112<rc(A)> [easy]
=1232031023022023000231213310101132232112T [easy]
=A1232031023022023000231213310101132232112
The last step is computationally hard, because it requires stepping through the sequence to work in reverse to find the first base.

And for the last trick, errors are fairly common in color-space sequence reads:
Code:
A2112322311010133121320003202203201302321 [original]
A2112322311010133121310003202203201302321 [error at position 21, before the 000]
The base-space representation of these sequences:
Code:
AGTGATCTACAACCATACTGCTTTTAGGAGGCTTGCCTAGT [original]
AGTGATCTACAACCATACTGCAAAATCCTCCGAACGGATCA [after a single error]
This error has caused the base-space representation to switch from decoded sequence 0 above to decoded sequence 3 at position 22. A match of the base-space representation of this sequence would have 20 nucleotide differences, while there is only a single difference in color-space. The great differences between base-space representations are why color-space sequences should be assembled and/or mapped in color-space.
gringer is offline   Reply With Quote