Bio Informatics and Machine Intelligence

1
Bio Informatics and Machine Intelligence

• Nicholas Flann
• Computer Science
• (based on a bioinformatics class of Larry Hunter)

2
Overview
3
Molecular biology databases

• There is a tremendous amount of information about
  bio-molecules in publicly available databases.
• Computational Tools are needed to efficiently
  find and process the data

4
Example of data growth
5
Data about Databases

• Nucleic Acids research publishes an annual
  database issue. 2003 issue lists 348 editorially
  selected databases
• Small excerpt from the A's
• AARSDB Aminoacyl-tRNA synthetase sequences
• ABCdb ABC transporters
• AceDB C. elegans, S. pombe, and human sequences
  and genomic information
• ACTIVITY Functional DNA/RNA site activity
• ALFRED Allele frequencies and DNA polymorphisms

6
What can be discovered about a gene by a database
search?

• Evolutionary information homologous genes,
  taxonomic distributions, etc.
• Genomic information chromosomal location,
  regulatory regions, shared domains, etc.
• Structural information associated protein
  structures, fold types, structural domains
• Expression information expression specific to
  particular tissues, developmental stages,
  phenotypes, diseases, etc.
• Functional information enzymatic/molecular
  function, pathway/cellular role, localization,
  role in diseases

7
Searching for informationabout genes and their
products

• Gene and gene product databases are often
  organized by sequence
• Genomic sequence encodes all traits of an
  organism.
• Gene products are uniquely described by their
  sequences.
• Similar sequences among bio-molecules indicates
  both similar function and an evolutionary
  relationship
• Macromolecular sequences provide biologically
  meaningful keys for searching databases

8
Searching sequence databases

• Many kinds of input sequences
• Could be amino acid or nucleotide sequence
• Genomic or mRNA/cDNA or protein sequence
• Complete or fragmentary sequences
• Goal is to retrieve a set of similar sequences

9
(No Transcript)
10
BLAST Searching with a sequence

• Goals is to find other sequences that are more
  similar to the query than would be expected by
  chance.
• Can start with nucleotide or amino acid sequence,
  and search for either (or both)

11
(No Transcript)
12
Example sequence

• atgcacttgagcagggaagaaatccacaaggactcaccagtctcctggtc
  tgcagagaagacagaatcaacatgagcacagcaggaaaagtaatcaaatg
  caaagcagctgtgctatgggagttaaagaaacccttttccattgaggagg
  tggaggttgcacctcctaaggcccatgaagttcgtattaagatggtggct
  gtaggaatctgtggcacagatgaccacgtggttagtggtaccatggtgac
  cccacttcctgtgattttaggccatgaggcagccggcatcgtggagagtg
  ttggagaaggggtgactacagtcaaaccaggtgataaagtcatcccactc
  gctattcctcagtgtggaaaatgcagaatttgtaaaaacccggagagcaa
  ctactgcttgaaaaacgatgtaagcaatcctcaggggaccctgcaggatg
  gcaccagcaggttcacctgcaggaggaagcccatccaccacttccttggc
  atcagcaccttctcacagtacacagtggtggatgaaaatgcagtagccaa
  aattgatgcagcctcgcctctagagaaagtctgtctcattggctgtggat
  tttcaactggttatgggtctgcagtcaatgttgccaaggtcaccccaggc
  tctacctgtgctgtgtttggcctgggaggggtcggcctatctgctattat
  gggctgtaaagcagctggggcagccagaatcattgcggtggacatcaaca
  aggacaaatttgcaaaggccaaagagttgggtgccactgaatgcatcaac
  cctcaagactacaagaaacccatccaggaggtgctaaaggaaatgactga
  tggaggtgtggatttttcatttgaagtcatcggtcggcttgacaccatga
  tggcttccctgttatgttgtcatgaggcatgtggcacaagtgtcatcgta
  ggggtacctcctgattcccaaaacctctcaatgaaccctatgctgctact
  gactggacgtacctggaagggagctattcttggtggctttaaaagtaaag
  aatgtgtcccaaaacttgtggctgattttatggctaagaagttttcattg
  gatgcattaataacccatgttttaccttttgaaaaaataaatgaaggatt
  tgacctgcttcactctgggaaaagtatccgtaccattctgatgttttgag
  acaatacagatgttttcccttgtggcagtcttcagcctcctctaccctac
  atgatctggagcaacagctgggaaatatcattaattctgctcatcacaga
  ttttatcaataaattacatttgggggctttccaaagaaatggaaattgat
  gtaaaattatttttcaagcaaatgtttaaaatccaaatgagaactaaata
  aagtgttgaacatcagctggggaattgaagccaataaaccttccttctta
  accatt

13
(No Transcript)
14
(No Transcript)
15
(No Transcript)
16
(No Transcript)
17
(No Transcript)
18
(No Transcript)
19
(No Transcript)
20
(No Transcript)
21
(No Transcript)
22
(No Transcript)
23
(No Transcript)
24
(No Transcript)
25
(No Transcript)
26
(No Transcript)
27
(No Transcript)
28
(No Transcript)
29
The complexity of database search

• Several algorithms for exact methods
• exhaustive (linear time, constant space)
• indexed (log time, linear space)
• hash tables (constant time, linear space)
• Inexact search methods
• Dynamic programming (Smith-Waterman)
• BLAST

30
Pairwise Sequence Alignment

• An alignment is a mapping from one sequence to
  another, identifying elements that are likely to
  have arisen from a common ancestor
• Alignments are NOT exact matches.

31
Similarity

• Similarity can be defined by counting positions
  that are identical between two sequences
• Gaps (insertions/deletions) can be important
  abcdef abcdef abcdef
  abceef acdef a-cdef

32
How many alignments?

• Without gaps, there are are NxM possible
  alignments between sequences of length N and M
• Once we start allowing gaps, there are many
  possible arrangements to consider abcbcd
  abcbcd abcbcd
  abc--d a--bcd ab--cd
• Every possible pairing between elements gives NM
  possible alignments.

33
Finding the optimal alignment

• Dynamic programming identifies optimal alignments
  in time proportional to the sum of the lengths of
  the sequences

34
Dynamic programming

• The key idea is to start aligning the sequences
  left to right once a prefix is optimally
  aligned, nothing about the remainder of the
  alignment changes the alignment of the prefix.
• We construct a matrix of possible alignment
  scores (NxM2 calculations worst case)
• Called Needleman-Wunch or Smith-Waterman

35
Alignment matrix

• Create a matrix with each sequence to be aligned
  along one edge and the score of the alignment of
  each pair of elements in a cell.
• Best local alignment is just the highest
  scoring diagonal

36
Dynamic programming matrix

• Each cell has the score for the best aligned
  sequence prefix up to that position.

37
The complexity of database search

• Several algorithms for exact methods
• exhaustive (linear time, constant space)
• indexed (log time, linear space)
• hash tables (constant time, linear space)
• Inexact search methods
• Dynamic programming (Smith-Waterman)
• BLAST

38
Pairwise Sequence Alignment

• An alignment is a mapping from one sequence to
  another, identifying elements that are likely to
  have arisen from a common ancestor
• Alignments are NOT exact matches.

39
Similarity

• Similarity can be defined by counting positions
  that are identical between two sequences
• Gaps (insertions/deletions) can be important
  abcdef abcdef abcdef
  abceef acdef a-cdef

40
How many alignments?

• Without gaps, there are are NxM possible
  alignments between sequences of length N and M
• Once we start allowing gaps, there are many
  possible arrangements to consider abcbcd
  abcbcd abcbcd
  abc--d a--bcd ab--cd
• Every possible pairing between elements gives NM
  possible alignments.

41
Finding the optimal alignment

• Dynamic programming identifies optimal alignments
  in time proportional to the sum of the lengths of
  the sequences

42
Dynamic programming

• The key idea is to start aligning the sequences
  left to right once a prefix is optimally
  aligned, nothing about the remainder of the
  alignment changes the alignment of the prefix.
• We construct a matrix of possible alignment
  scores (NxM2 calculations worst case)
• Called Needleman-Wunch or Smith-Waterman

43
Alignment matrix

• Create a matrix with each sequence to be aligned
  along one edge and the score of the alignment of
  each pair of elements in a cell.
• Best local alignment is just the highest
  scoring diagonal

44
Dynamic programming matrix

• Each cell has the score for the best aligned
  sequence prefix up to that position.

45
Why predict protein structure?

• Neither crystallography nor NMR can keep pace
  with genome sequencing efforts
• Computer scientists love this problem
• Understandable with minimal biology
• Seems like a good discrimination task
• Understand the mechanisms of folding (?)
• First computational Nobel prize?

46
Protein structure

• Amino acid sequence alone should be enough to
  determine protein structure
• However, the physics are daunting
• 20,000 protein atoms, plus equal amounts of
  water
• Can takes seconds (most chemical reactions take
  place 1012 --1,000,000,000,000x faster)
• Empirical determinations of protein structure are
  advancing rapidly

47
Protein structure cartoons
48
Protein Structure Representations

• Differentvisualizationsshow variousaspects
  ofstructure

49
Competition

• Critical Assessment of Structure Prediction
  competition since 1994
• Solved, but unpublished structures are posted in
  May, predictions due in September
• Various categories
• Relation to existing structures, ab initio,
  homology, fold, etc.
• Partial vs. Fully automated approaches
• Produces lots of information about what aspects
  of the problems are hard, and ends arguments
  about test sets.
• Results showing steady improvement, and the value
  of integrative approaches.

50
Reconstructing Phylogenies

• The only universal biological fact is that all
  species are related by descent.
• The genealogical history of life can be
  represented by an evolutionary tree (or
  phylogeny), with contemporary organisms at the
  leaves.
• Goal of phylogenetic inference is to reconstruct
  the order of splitting events (and perhaps the
  distances between them).

51
The phylogeny problem

• Input
• A set of contemporary species (S) whose
  evolutionary relationship is to be reconstructed
• A set of inheritable characteristics (C) that
  describe each species. Characteristics can be
• quantitative (continuous, e.g. size)
• qualitative (discrete, e.g. gene sequence)
• Output
• Tree (perhaps branch lengths) which best fits the
  data.
• Assumptions
• Common ancestor

52
Why do it?

• Resolve evolutionary history
• Constructing vaccines
• Want to assure that vaccine is constructed to
  address diverse strains of the disease (e.g.
  influenza)
• Epidemiology
• Reconstruct paths of infection, either of an
  individual or generally. (e.g. Crandall's
  Evolution of HIV)

53
Tree reconstruction as optimization

• Number of possible trees is 2S. Considering
  branch lengths makes the problem harder.
• Want the tree that maximizes some quality score.
  Score based on either
• Characters (e.g. sites in a sequence) directly,
  or
• Distances between character sets (e.g. alignment
  scores)
• Another NP-hard optimization problem...

54
Managing the Biomedical Literature

• Challenges are huge
• 700,000 articles/yearaccelerating 10
  year!16,048 last week
• Large number of relevant articles for most topics
• (Sub-)Disciplinary boundaries are breaking down
• Computational tools are crucial
• Information retrieval (finding the right
  articles)
• Organization dissemination of articles data
• Information extraction

55
Natural Language Processing

• NLP is many tools and technologies.
• Natural language understanding capture the
  meaning and implications of texts. A mostly
  unobtainable goal, but much useful theory
• Using documents as evidence for gene function or
  other kinds of inference
• Information extraction capture the meaning and
  implications of certain restricted aspects of
  text
• Document management reasoning about documents,
  rather than the texts they contain
• Information retrieval (including Q/A)
• Browsing and organizing large collections of
  documents

56
Summary

• Many Significant and Difficult Computational
  Problems
• Need Computer Scientists to Develop new efficient
  algorithms and methodologies
• Many different fields
• Graphics, Algorithms, Machine Intelligence,
  Machine Vision, Optimization, Databases, etc.