The MoBIoS Project Molecular Biological Information System

1
The MoBIoS ProjectMolecular Biological
Information System

• Daniel P. Miranker
• Dept. of Computer Sciences
• Center for Computational Biology and
  Bioinformatics
• University of Texas

Weijia Xu, Rui Mao, Will Briggs, Smriti
Ramakrishnan, Shu Wang, Shulin Ni, Kai Yan, Ving
Lei
2
Compared to Business Databases,Biological
Databases

• are not that big
• Genbank ftp mirror download, ltlt 1 Terabyte
• CMS spectrometer at CERN, gt Petabyte/year
• but,
• data management is biology is a big problem
• There must be another problem

3
You Cant Sort

• Sequences
• DNA, RNA, Protein databases
• Mass Spectra
• proteomics
• Small Molecules Protein Structure
• Protein interaction
• Rational drug design
• Pathways (graphs)
• Phylogenies (graphs, trees in particular)

4
In Life-Sciences Database Management Systems are
Souped Up File Systems

• Primary data is stored in text or blob fields
• Annotations may be relational

• Data retrieval
• Filter DB, sequential dump, O(n), to utilities
• E.g. BLAST,

5
Scope To Find Common Ground Both Biology and
DBMS Have to Move
DBMS
Biological Information System
Metric-Space Database as the Common Ground
6
Metric Space is

• a pair, M(D,d),
• where
• D is a set of points
• d is metric distance function with the
  following properties
• d(x,y) d (y,x)
  (symmetry)
• d(x, y) gt 0, d(x,x) 0
  (non negativity)
• d(x,z) lt d(x,y) d(y,z)
  (triangle inequality)

x y z
7
Definition - By Analogy

• A Spatial Database Management System
• Extend relational DBMS
• Special indexes for 2D and 3D data k-d and
  R-trees
• New data types
• Geographic information systems
• Topographic maps
• Buildings and the like

• A Metric-Space Database Management System
• Extend Relational DBMS
• Special indexes for metric-spaces
• New data types
• Biological information system
• Life science data types

8
Results to date

• Developed and Validated Metrics for
• protein sequence homology
• protein mass-spectra
• volumetric (3-d) models of protein electrostatics
• Feasibility prototype,
• validated scalable biological data retrieval

9
A Broad Range of Problems

• General purpose metric-space index
• Use of metrics as similarity functions on
  biological data
• Integration of biological sequences into the SQL
  programming model

10
Metric Space Indexing
Vantage point algorithm BurkhardKeller73
Choose a point,VP
And a radius, R
11
Range Search
R

• If d(q,VP) gt R d
• then
• search outside the sphere

12
Biological Models are Usually Based on Similarity

• Similarity
• Biologists like scoring functions that reward
  each similar feature with a positive number
• Intuitive
• Distance
• More similar ? smaller numbers
• Identical ? 0

13
1) Do Metric Models of Similarity Capture
Biology?

• Metrics are a subset of possible mathematical
  models

.
14
Sequence Similarity

• Sequence similarity based on weighted edit
  distance
• Sellers74
• Accepted weight matrices, PAM BLOSSUM, are not
  metric
• Log-odd matrices negative values
• Defy simple algebraic normalization
• TaylorJones93,Linialetal97, Halperin etal04

15
PAM vs. mPAM t 1/f
Xu Miranker04

• Using original substitution counts
• PAM frequency of substitution
• S(a,bt) log P(ba,t)/qb
• mPAM expected time between substitutions
• D(a,b) 1/log(1 ?(P(a,x)P(b,x))

x
16
Metrics for Biological Similarity

• Biosequences
• mPAM
• mBLOSSUM (not yet published)
• Hamming distance
• Protein Mass and Protein Fragment Mass Spectra
• derived an effecive metric from cosine distance
• Volumetric models of chemical fields of proteins
• (with C. Bajaj)

17
Matching Electrostatic Shape of Molecules
18
Metric-Space Indexing (search)

• Well studied in main-memory
• no means a closed problem
• In databases (external/disk based methods)
• Embryonic
• Success only in specific domains
• Biological sequences, nucleotide and peptide
• Databases of analytically determined mass-spectra
• Volumetric models of proteins (collaboration in
  progress with C. Bajaj)

19
MoBIoS System Overview
Other Opportunities
Active Application Efforts
Other Opportunities
Active Application Efforts
Mass
-
Spec
Mass
-
Spec
Combi
-
Chem
Combi
-
Chem
Compartive
and
Compartive
and
Homology
Homology
Ligand
Ligand
Library
Protein
Library
Protein
Search
Search
Phylo
Genomics
Phylo
Genomics
Docking
Docking
Management
Identification
Management
Identification
MoBIoS
SQL (mSQL)
-
)
MoBIoS
SQL (
Query Engine
Mining Engin
Query Engine
Mining Engine
MoBIoS
Java Interface (MJI)
Metric
-
Space Based
Metric
-
Space Based
Storage Manager
Storage Manager
20
SQL as a Bioinformatic Programming Model

• Simple retrieval of similar objects.

21
Protein identification by database look-up

• Build a reference database
• Given a complete genome ? putative proteome
• Given a proteome ? set of ideal mass-spectra
  (database)
• 2. Wet lab work
• In a laboratory, experimentally determine the
  mass-spectra of an unknown protein
• 3. Analyze the data
• Find the closest match with the database
• (a very noisy proposition)

22
Analytic Mass-Spectra

• At each resolvable mass, either a peptide is
  present or not
• Vector-Space model binary vector, one bit for
  each resolvable mass
• Similarity Shared peaks count Inner Product
• (0100101) (0111100) 2

23
Cosine Distance

• Drs 1 xrxs/(xrxr)1/2(xsxs)1/2
• Document retrieval uses
• the vector space model
• known
• similarity inner product
• as distance cosine distance

24
Given - the spectra of an unknown, S -
reference database, Spectra

• SELECT Spectra.accesionNumber
• FROM Spectra
• WHERE
• Cosine_Distance(S,Spectra,tolerance)

25
Protein Identification by Database Lookup of
Mass Spectra
26
Sequences Pose Additional Problems

• Sequences long units (identity for storage and
  retrieval)
• Genes
• Chromosomes
• Analysis comprises comparing small substrings

30/45
27
Local-Alignment (Homology Search)

• For each pair of arguments
• Find matching subsequences

Sequence 1
Sequence 2
28
Soln Sequence View

• New view type
• Breaks sequences into k-mers

create SEQUENCEVIEW rice_sview as SELECT CREATE
FRAGMENTS (, 3) k 3 FROM WHERE
USING HAMMING-DISTANCE Dan Instead of
miRNA later, maybe Sequence view of only exons.
29
Materialize as an Index
D(AAA) 2


30
mSQL Sequence Operators

• CreateFragement
• Merge and Groupby
• Merge fragments back into longer subsequences
• Algebraic constraints
• Consider relative offsets when applied to pairs
  of sequences
• Parameterize this for additional biological
  semantics (gaps)

31
Local-Alignment (Homology Search)

• For each pair of arguments
• Find matching subsequences

Sequence 1
Sequence 2
32
Dan Insert example of n log n homology on k-mers
with merge
33
Conserved (PCR) Primer Pair Discovery(with R.
Linder)

• Goal
• To find evidence of horizontal gene transfer
• Method
• Sample and sequence many orthologous regions in
  many plants
• Build a phylogenetic tree for each region, look
  for inconsistent topology

34
Pattern Definition of a Conserved Primer Pair

• Compare Arabidopsis Genome X Rice Genome
• Locate nucleotide patterns of form
• PCR primer pair candidate
• Matching ? lt 5 mismatches
• Eliminate non-unique primer candidates
• Usual implementations O(n2), n 109

Rice Arab.
?18 Matching Nucleotides
?18 Matching Nucleotides
Rice Gap 400 3000 Long Arab. Gap 400 3000
Long
35
Query Plan

• Arab. Genome, O(n)
  Rice Genome, O(m)
• Offline Build Sequence
• View O(n log n)
• Compare O(mlogn)
• Indexed Nested Loop
• Eliminate Duplicates
• Eliminate Low Complexity
• Primers (LZ compression)
• Merge Overlapping Primers
• 10,000 conserved
• primer pairs candidates

36
mSQL to locate candidate conserved primer pairs
// Create sequenceview of 18-mers of Rice and
Arab.

• SELECT merge(R1.fragment, A1.fragment)
• FROM
• Rice_sview R1, Rice_sview, R2, Arab_sview A1,
  Arab_sview A2
• WHERE
• distance(HAMMINGDISTANCE', R1.fragment,
  A1.fragment) lt 1.0 AND distance(HAMMINGDISTANCE'
  , R2.fragment, A2.fragment) lt 1.0 AND
• (FRAGOFFSET(R2.fragment)-FRAGOFFSET(R1.fragment))
  gt 400 AND
• (FRAGOFFSET(R2.fragment)-FRAGOFFSET(R1.fragment))
  lt 3000 AND
• (FRAGOFFSET(A2.fragment)-FRAGOFFSET(A1.fragment))
  gt 400 AND
• (FRAGOFFSET(A2.fragment)-FRAGOFFSET(A1.fragment))
  lt 3000
• GROUP BY R1.fragment, A1.fragment

37
Detailed Analysis, Wet Lab Validation

• Found 13,418 possible primer pairs from MoBIoS,
• lt 8 processor days including database load
• 100 best candidates BLASTed for matches in
  GenBank
• 15 matched other plant genes and the primers
• At least 2 of 15 showed potential after PCR
  amplification against Helianthus and
  Phalaenopsis.

38
mSQL Queries Developed for Other Sequence
Analysis Problems Miranker et.al. Data Eng.
Bul. Sept. 04

• Micro RNA RNAi predictions
• Rosetta analysis to determine protein function
• Electronic-PCR
• Sequence Homology

39
Query Engine Can Consider All Data

• From the MiRNA problem
• Select merge(G7.dna_seq), merge(M7.seq) mseq,
  gene
• From M7, G7, Features // consider both 7-mers
  and their annotations
• Where
• distance(hamming, G7.dna_seq, M7.seq) lt 0
• and G7.SID Features.SID
• and Features.name miRNA
• and G7.dna_seq.offset gt Features.first
• and G7.dna_seq.offset G7.dna_seq.length lt
  Features.last
• having mseq.offset lt 1 and mseq.offset
  mseq.length gt 7 

40
Status

• All components of the system integrated and
  function
• Not product
• Scalability of index on nearest-neighbor search
• Range search is a better fit for the relational
  model
• Actively seeking collaborations in anticipation
  of distributing the software.

41
Protein Identification by Database Lookup of
Mass Spectra
42
Matching Electrostatic Shape of Molecules
43
Status

• Started with McKoi
• A Java open source object-relational DBMS
• (Think of Postgress written in Java)
• Added
• Biological data types
• Metric-space index
• Extending SQL engine (in finishing stages)

44
Other Results

• Protein identification by database lookup
• match against a database of theoretical
  mass-spectrometer signatures.
• Protein similarity by 3-D electrostatic shape

45
Three classes of algorithm

• Vantage-Point
• Generalized Hyperplane
• Radius-based Methods

46
Hyper-planes Ulhmann91

• If d(x,h1) lt d(x,h2) then x assigned to h1

h1
x
h2
47
Develop a Hierarchical Clustering
C
A
E
B
D
F

• Hierarchy of Bounding spheres, (center, radius),
• Bounding spheres may overlap
• Inspired by R-trees

48
Multi-vantage point method

• Consider d(VPi, x) a projection onto an axis
• Looks like a k-d tree
• Choose number k d

49
Hyper-planes Ulhmann91

• If d(x,h1) lt d(x,h2) then x assigned to h1

h1
x
h2
50
Status

• Implemented an algorithm from each class.
• Examined performance on
• Synthetic euclidean vectors typical of other work
• Peptide k-mers, mPAM distance
• Mass spectra, fuzzy cosine distance
• Image database

51
Comparison of three methods

• Vector data not typical of other data sets

• Multivantage point wins
• but not yet exciting

52
Our First Result mPAM XuMiranker04

• Dayhoff etals PAM Derivation74
• Took a set of closely related protein sequences
• Developed a phylogenetic tree based on parsimony
• Counted substitutions to transform one sequence
  to another

53
mPAM250 vs. PAM250 on NCBI benchmark

• 103 queries over yeast proteome
• Hand curated correct answers.