Initial Proposal for the RNA Alignment Ontology

1
Initial Proposal for the RNA Alignment Ontology

• Rob Knight
• Dept Chem Biochem
• CU Boulder

2
What do we want to do?

• Represent detailed structural info and other
  metadata on alignment
• Avoid horizontal and vertical expansion
• Explicitly annotate correspondences at the level
  where they occur

3
What do alignments look like now?
4
Why is this a problem?
5
so real alignments look like this, to shoehorn
everything into columns that are assumed to be
homologous
6
Homology is problematic

• Fundamental problem systems that are homologous
  at one level are not necessarily homologous at
  other levels
• E.g. bat wings and bird wings homologous as
  pentadactyl limbs, but not homologous as wings
• Homology is hierarchical andcan partially
  overlap at any level(e.g. Griffiths 2006)

Bat forelimbs
Bird forelimbs
Frog forelimbs
Rodent forelimbs
Mammal forelimbs
Tetrapod forelimbs
Ridley Evolution 3rd ed.
7
and correspondence need not be homology at all!

• Example from SELEX hammerhead ribozymes
  independently evolved at least three times in
  nature, and in Jack Szostak and Ron Breakers
  labs
• However, we still want to be able to align the
  functionally equivalent sequences although there
  is not evolutionary relationship

8
So what are going to use the alignment ontology
for?
9
Use case 1 aligning rRNA
10
Problem have millions of fragments, want to
align (incl. noncanonical pairs) assign
named regions
11
Solution

• Use existing alignment, try to fit new seqs in
• Would be improved if we could explicitly annotate
  helices, noncanonical pairs, etc. on the sequence
  overall
• For display, need to easily show/hide groups of
  sequences and/or regions of the sequence

12
Use case 2 SELEX

• From large number of unaligned sequences, want to
  identify motifs like this (Majerfeld Yarus 2005)

13
How is this currently done?

• Find regions that are similar in more sequences
  than chance
• Group these sequences centered on the motif
• See if the parts of the motif can be related by
  helices
• See if anything else is reliably found by the
  motif
• Repeat for other families and see if there are
  relationships between them
• Group these families together, then iterate

14
e.g. here we discovered unpaired G important
15
So how do we handle all this? A proposal

• Entities
• sequence_region a thing that defines a set of
  bases relative to some sequence (i.e. with
  indices for each base)
• paired_sequence_region two regions linked by
  pairs
• helical_sequence_region two regions completely
  paired
• base region that consists of single nucleotide
• base_pair region that consists of two, paired
  bases
• canonical_base_pair base pair that is cis-WW
• loop contiguous sequence_region stretching from
  i to j such that i-1 and j1 are a base pair
• etc. (bulge, internal_loop, junction, etc.)

16
So how do we handle all this? A proposal

• Relationships
• correspondence relation among set of
  sequence_regions implying all share a feature
  (with metadata about how determined)
• homology correspondence implying continuous
  chain of descent preserving the relation
• sequence_similarity correspondence implying
  regions are similar in primary sequence
• two_d_structure_similarity correspondence
  implying regions are similar in 2D structure,
  i.e. nested canonical base pairs
• secondary_structure_similarity correspondence
  implying regions are similar in secondary
  structure, i.e. incl. pseudoknots/noncanonicals
• tertiary_structure_similarity correspondence
  implying regions are similar in 3D structure

17
So how do we handle all this? A proposal

• Relationships
• pairing relation that asserts that two
  sequence_regions each have parts of at least one
  base_pair that connects them
• helical_pairing pairing that includes several
  base_pairs (not necessarily contiguous) between
  two sequence_regions
• unbroken_helical_pairing helical_pairing that
  includes no bases in the sequence_regions that
  are not paired with the other sequence_region, in
  order
• base_pairing pairing that connects exactly two
  bases, annotated with the Leontis-Westhof
  classification
• More exotic uses for alignment
• microrna_target pairing relation in which one
  member is a miRNA and the other is an mRNA
  according to SO
• same_microrna_target a relation among a set of
  sequences that have microrna_target relation to
  the same miRNA

18
Implementation notes

• Must be able to name regions (e.g. P3 in RNaseP)
  and subclass them (e.g. P3 in firmicutes)
• Must be able to subclass homologies, e.g.
  homologous as wing vs. homologous as limb
• Correspondences are all symmetric and transitive,
  so can implement as set of regions that share the
  correspondence
• (probably) dont want to reify names of parts of
  well-known RNAs in the overall RNAO?

19
Acknowledgements

• RNA Alignment Ontology working group
• James. W. Brown
• Fabrice Jossinet
• Rym Kachouri
• B. Franz. Lang
• Neocles Lenotis
• Gerhard Steger
• Jesse Stombaugh
• Eric Westhof
• Other coauthors
• Amanda Birmingham
• Paul Griffiths
• Franz Lang

• Knight Lab members
• Cathy Lozupone
• Micah Hamady
• Chris Lauber
• Jesse Zaneveld
• Jeremy Widmann
• Elizabeth Costello
• Jens Reeder
• Daniel McDonald
• Anh Vu
• Ryan Kennedy
• Julia Goodrich
• Meg Pirrung
• Reece Gesumaria

• Trp project
• Irene Majerfeld
• Jana Chochosolousova
• Vikas Malaiya
• Matthew Iyer
• Mike Yarus

NSF RCN grant 0443508