MicroRNA identification based on sequence and structure alignment

1
MicroRNA identification based on sequence
andstructure alignment
Xiaowo Wang, Jing Zhang, Fei Li, Jin Gu, Tao
He, Xuegong Zhang and Yanda Li

• Presented by -
• Neeta Jain

2
Outline

• Introduction
• Motivation
• Experiment
• Materials
• Methods
• Results
• Conclusion

3
Introduction

• What are miRNAs and why are they important?
• miRNAs are 22 nt long non-coding RNAs
• They are derived from their 70 nt precursors,
  which typically have a hairpin structure

• Importance of miRNAs
• They are found to regulate the expression of
  target genes via complementary base pair
  interactions.

4
Motivation

• Since miRNAs are short (22 nt), conventional
  sequence alignment methods can only find
  relatively close homologues
• It has been reported that miRNA genes are more
  conserved in their secondary structure than in
  primary structure
• This paper exploits this secondary structure
  conservation and proposes a novel computational
  approach to detect miRNAs based on both sequence
  and structure alignment
• The authors devised a tool miRAlign and have
  compared its performance with existing searching
  methods such as BLAST and ERPIN

5
Experiment

• Materials
• Reference sets
• Consists of 1298 miRNAs from 12 species out of
  which 1054 were animal miRNAs.
• 1054 animal miRNAs and their precursors(1104)
  composed our raw training set Train_All.
• Train_Sub_1 All animal miRNAs except those from
  C.briggsae
• Train_Sub_2 All animal miRNAs except those from
  C.briggsae and C.elegans
• Genomic sequences
• Sequences of 6 species were used.

6

• Methods
• Preprocessing
• Known precursors from training set are used to
  BLAST against the genome
• Potential regions are cut from the genome with 70
  nt flanking sequences to each end
• Such regions are scanned using a 100nt window
  with 10 nt step
• Overlapping sequences with repeat sequences are
  discarded.

7

• Methods (contd)
• miRAlign
• Secondary Structure Prediction
• Both the candidate sequence and its reverse
  complement are analyzed by RNA fold to predict
  hairpins.
• Only hairpins with MFE lower than -20 kcal/mol
  are retained.
• Pairwise sequence alignment
• Sequences from previous step are aligned pairwise
  to all the 22 nt known miRNA sequences from the
  training set
• Sequence similarity score between the candidate
  and known mature miRNAs is calculated by
  CLUSTALW.
• If the score exceeds a user-defined threshold,
  then the candidate to known miRNA pairs are kept
  for further analysis

8

• Methods (contd)
• Checking miRNAs position on stemloop
• 3 properties for miRNAs position are considered
• Should not locate on terminal loop of hairpin
• Should locate on the same arm of hairpin
• Position of potential miRNA on hairpin should not
  differ too much from its known homologues
• Position difference of miRNA on precursors A and
  B

9

• Methods (contd)
• RNA secondary structure alignment
• RNAforester computes pairwise structure alignment
  and gives similarity score
• Score is a summation of all base (base pair)
  match (insertion, deletion).
• Normalized similarity score of structure C and m
  is given as

where, C Candidate sequence m known
pre-miRNA sigma_local(C,m) raw local alignment
score between C and m Sigma(m,m) self-alignment
score of m
10

• Methods (contd)
• Total similarity score
• After aligning all potential homologue pairs, a
  total similarity score (tss) is assigned to each
  candidate sequence.

Where, C- candidate sequence R set composed
of all Cs
11
Methods (contd)
Summary -
12
Results

• Application on C.briggsae
• Detection of miRNA homologues -
• miRAlign was applied on C.briggsaes data with
  training
• set Train_Sub_1 and sensitivity and specificity
  were recorded.
• Identification of miRNAs in distantly related
  species -
• miRAlign was applied on C.briggsaes data with
  training
• set Train_Sub_1 and sensitivity and specificity
  were recorded

13
Results (contd)
Graph 1 -
14
Results (contd)
Graph 2 -
15
Results (contd)
Comparison of miRAlign with BLAST -
16
Results (contd)
Comparison of miRAlign with ERPIN -
17
Results (contd)

• Other results
• miRAlign was applied to A. gambiae and 59
  putative miRNAs with tss gt 35 were detected .
  This was validated when 38 A. gambiae miRNAs were
  reported in the MicroRNA registry 6.0 and 37 of
  them were covered by miRAlign
• miRAlign was also applied to plant, Zea mays and
  detected 28 out of 40 known Zea Mays miRNAs.

18
Conclusion

• Combining sequence and structure alignments,
  miRAlign has better performance than previously
  reported homologue search methods
• Although, mirAlign was based on animal data, the
  miRNAs predicted in Zea mays indicates that
  miRAlign can be applied to plants. Further
  investigation regarding this is underway.

19
THANK YOU
Questions ??