Transcript CSCE590/822 Data Mining Principles and Applications

CSCE555 Bioinformatics
Lecture 9 Gene Finding & Comparative
genomics
HAPPY CHINESE NEW YEAR
Meeting: MW 4:00PM-5:15PM SWGN2A21
Instructor: Dr. Jianjun Hu
Course page: http://www.scigen.org/csce555
University of South Carolina
Department of Computer Science and Engineering
2008 www.cse.sc.edu.
Outline
Performance Evaluation of Gene Finding
programs
 Comparative genomics:

◦
◦
◦
◦
What to do
Tools
Databases
Application case
7/15/2015
2
Accuracy Measures of Gene-Finding Programs
Sensitivity vs. Specificity (adapted from Burset&Guigo 1996)
TP
FP
TN
FN
TP
FN
TN
Actual
Predicted
Actual
No Coding / Coding
Predicted
Coding / No Coding
TP
FP
FN
TN
Sensitivity
(Sn)
Fraction of actual coding regions that are correctly predicted as
coding
Specificity
(Sp)
Fraction of the prediction that is actually correct
Correlation
Coefficient (CC)
Combined measure of Sensitivity & Specificity
Range: -1 (always wrong)  +1 (always right)
Test Datasets

Sample Tests reported by Literature
◦ Test on the set of 570 vertebrate gene seqs
(Burset&Guigo 1996) as a standard for
comparison of gene finding methods.
◦ Test on the set of 195 seqs of human, mouse
or rat origin (named HMR195) (Rogic 2001).
Results: Accuracy Statistics
Table: Relative Performance (adapted from Rogic 2001)
Complicating Factors for Comparison
• Gene finders were trained on data
that had genes homologous to test
seq.
• Percentage of overlap is varied
• Some gene finders were able to
tune their methods for particular data
# of seqs - number of seqs effectively analyzed
by each program; in parentheses is the number
of seqs where the absence of gene was
predicted;
Sn -nucleotide level sensitivity; Sp - nucleotide
level specificity;
CC - correlation coefficient;
ESn - exon level sensitivity; ESp - exon level
specificity
• Methods continue to be developed
Needed
• Train and test methods on the same
data.
• Do cross-validation (10% leave-out)
GenScan compared to other genefinding programs
Why not Perfect?

Gene Number
usually approximately correct, but may not

Organism
primarily for human/vertebrate seqs; maybe lower accuracy for
non-vertebrates. ‘Glimmer’ & ‘GeneMark’ for prokaryotic or
yeast seqs

Exon and Feature Type
Internal exons: predicted more accurately than Initial or Terminal
exons;
Exons: predicted more accurately than Poly-A or Promoter
signals

Biases in Test Set (Resulting statistics may not be representative)
Eukaryotic Gene Finding Tools

Genscan (ab initio), GenomeScan (hybrid)

(http://genes.mit.edu/)

Twinscan (hybrid)

(http://genes.cs.wustl.edu/)

FGENESH (ab initio)

(http://www.softberry.com/berry.phtml?topic=gfind)

GeneMark.hmm (ab initio)

(http://opal.biology.gatech.edu/GeneMark/eukhmm.cgi)

MZEF (ab initio)

(http://rulai.cshl.org/tools/genefinder/)

GrailEXP (hybrid)

(http://grail.lsd.ornl.gov/grailexp/)

GeneID (hybrid)

(http://www1.imim.es/geneid.html)
Comparative Genomics
Outline for Comparative Genomics
Overview
 Why do comparative genomic analysis?
 Assumptions/Limitations
 Genome Analysis and Annotation Standard Procedure
 General Purposes Databases for Comparative
Genomics
 Organism Specific Databases
 Genome Analysis Environments
 Genome Sequence Alignment Programs
 Genomic Comparison Visualization Tools

What is comparative genomics?
Analyzing & comparing genetic material
from different species to study evolution,
gene function, and inherited disease
 Understand the uniqueness between
different species

What is compared?
Gene location
 Gene structure

◦
◦
◦
◦

Exon number
Exon lengths
Intron lengths
Sequence similarity
Gene characteristics
◦ Splice sites
◦ Codon usage
◦ Conserved synteny
Figure 1 Regions of the human and mouse homologous genes: Coding exons
(white), noncoding exons (gray}, introns (dark gray), and intergenic regions
(black). Corresponding strong (white) and weak (gray) alignment regions of GLASS
are shown connected with arrows. Dark lines connecting the alignment regions
denote very weak or no alignment. The predicted coding regions of ROSETTA in
human, and the corresponding regins in mouse, are shown (white) between the
genes and the alignment regions.
Sequenced prokaryotic genomes
Bacteroides fragilis
Bordetella bronchiseptica
Bordetella parapertussis
Bordetella pertussis
Burkholderia cepacia
Burkholderia pseudomallei
Chlamidophila abortus
Clostridium botulinum
Clostridium difficile
Corynebacterium diphtheriae
Erwinia carotovora
Escherichia/Shigella spp. (5)
Mycobacterium bovis
Mycobacterium marinum
Neisseria meningitidis (serogroup C)
Salmonella typhi
Salmonella spp. (5)
Staphylococcus aureus (MRSA)
Staphylococcus aureus (MSSA)
Streptococcus pneumoniae
Streptococcus pyogenes
Streptococcus suis
Streptococcus uberis
Streptomyces coelicolor
Tropheryma whipelli
Wolbachia (Culex quinquefasciatus)
Wolbachia (Onchocerca volvulus)
Yersinia enterocolitica
Yersinia pestis
Opportunistic
Veterinary
Whooping cough
Whooping cough
Lung infections in CF
Melliodosis
Veterinary
Botulism
Colitis
Diphtheria
Plant pathogen
Various
Tuberculosis
Various
Bacterial meningitis
Typhoid fever
Various
Various (Nosocomial)
Various (Community acquired)
Bacterial meningitis
Various (ARF-associated)
Veterinary
Veterinary
Non-pathogenic
Whipple’s disease
Vector (Bancroftian filariasis)
River Blindness
Food poisoning
Plague
In progress
In progress
Complete
Complete
In progress
In progress
Funded
Funded
In progress
Complete
Funded
In progress
In progress
In progress
In progress
Complete
In progress
Complete
In progress
In progress
In progress
In progress
In progress
Complete
In progress
In progress
Funded
In progress
Complete
Sequenced eukaryotic genomes
Aspergillus fumigatus
Dictyostelium discoideum
Entamoeba histolitica
Leishmania major
Plasmodium falciparum
Schistosoma mansoni
Schizosaccharomyces pombe
Theileria annulata
Toxoplasma gondii
Trypanosoma brucei
Farmer’s lung
Soil amoeba
Amoebic dysentry
Leishmaniasis
Malaria
Bilharzia
Fission yeast
Veterinary
Toxoplasmosis
Sleeping sickness
In progress
In progress
In progress
In progress
In progress
In progress
Complete
In progress
In progress
In progress
Bioinformatics Flow Chart
1a. Sequencing
1b. Analysis of nucleic acid
seq.
6. Gene & Protein expression data
7. Drug
screening
2. Analysis of protein seq.
3. Molecular structure prediction
Ab initio drug design OR
Drug compound screening in
database of molecules
4. molecular interaction
8. Genetic variability
5. Metabolic and regulatory networks
Genome Sequencing Process
Genomic DNA
Shearing/Sonication
Subclone and Sequence
Shotgun reads
Assembly
Contigs
Finishing read
Finishing
Complete sequence
Genome Sequencing - Review
Strategy
Clone by clone vs whole genome shotgun
Libraries
Subcloning; generate small insert libraries
Sequencing
Assembly
Closure
Annotation
Release
•Most genome will be sequenced and can be sequenced;
few problem are unsolvable.
Assembly: Process of taking raw single-pass reads into
contiguous •Problem
consensus
sequence
(Phred/Phrap)
lies
in understanding
what you have:
Closure: Process of ordering and merging consensus
•Gene
finding
sequences into a
singleprediction/gene
contiguous sequence
•Annotation
-DNA features (repeats/similarities)
-Gene finding
Release
to the public e.g. EMBL or GenBank
-Peptidedata
features
-Initial role assignment
-Others- regulatory regions
Annotation of eukaryotic genomes
Genomic DNA
ab initio gene
prediction
transcription
Unprocessed RNA
RNA processing
Mature mRNA
Gm3
AAAAAAA
translation
Nascent polypeptide
Comparative gene
prediction
folding
Active enzyme
Functional
identification
Function
Reactant A
Product B
Why do comparative genomics?

Many of the genes encoded in each genome from the genome
projects had no known or predictable function

Analysis of protein set from completely sequenced genomes

Uniform evolutionary conservation of proteins in microbial
genomes, 70% of gene products from sequenced genomes have
homologs in distant genomes (Koonin et al., 1997)

Function of many of these genes can be predicted by comparing
different genomes of known functional annotation and transferring
functional annotation of proteins from better studied organisms to
their orthologs in lesser studied organisms.

Cross species comparison to help reveal conserved coding regions

No prior knowledge of the sequence motif is necessary

Complement to algorithmic analysis
Assumptions/Limitation
Homologous genes are relatively well preserved
while noncoding regions tend to show varying
degrees of conservation. Conserved noncoding
regions are believed to be important in
regulating gene expression, maintaiing structural
organization of the genome and most likely
other possible functions.
 Cross species comparative genomics is
influenced by the evolutionary distance of the
compared species.

Genome Analysis and Annotation: General Procedure
Basic procedure to determine the functional and structural annotation of
uncharacterized proteins:

Use a sequence similarity search programs such as BLAST or FASTA to identify all
the functional regions in the sequence. If greater sensitivity is required then the
Smith-Waterman algorithm based programs are preferred with the trade-off
greater analysis time.

Identify functional motifs and structural domains by comparing the protein
sequence against PROSITE, BLOCKS, SMART, CDD, or Pfam.

Predict structural features of the protein such as signal peptides, transmembrane
segments, coiled-coil regions, and other regions of low sequence complexity

Generate a secondary and tertiary (if possible) structure prediction

Annotation:
◦ Transfer of function information from a well-characterized organism to a lesser studied
organism and/or
◦ Use phylogenetic patterns (or profiles) and/or
◦ Use the phylogenetic pattern search tools (e.g. through COGs) to perform a systematic
formal logical operations (AND, OR, NOT) on gene sets -- differential genome display
(Huynen et al., 1997).
Automated Genome Annotation
GeneQuiz – limited number of searches/day
 MAGPIE – outside users cannot submit own
seq
 PEDANT – commercial version allow for full
capacity
 SEALS – semi automated

General Databases Useful for
Comparative Genomics








Locus Link/RefSeq: http://www.ncbi.nih.gov/LocusLink/
PEDANT -Protein Extraction Description ANalysis Tool
http://pedant.gsf.de/
MIPS – http://mips.gsf.de/
COGs - Cluster of Orthologous Groups (of proteins)
http://www.ncbi.nih.gov/COG/
KEGG - Kyoto Encyclopedia of Genes and Genomes
http://www.genome.ad.jp/kegg/
MBGD - Microbial Genome Database
http://mbgd.genome.ad.jp/
GOLD - Genome OnLine Database
http://wit.integratedgenomics.com/GOLD/
TOGA – http://www.tigr.org/xxxxx
Problems with existing sequence alignments
algorithms for genomic analysis




Most algorithms were developed for comparing single protein
sequences or DNA sequences containing a single gene
Most algorithms were based on assigning a score to all the possible
alignments (usually by the sum of the similarity/identity values for
each aligned residue minus a penalty for the introduction of gaps)
and then finding the optimal or near-optimal alignment based on
the chosen scoring scheme.
Unfortunately, most of these programs cannot accurately handle
long alignments.
Linear-space type of Smith-Waterman variants are too
computationally intensive requiring specialized hardware (memorylimited) or very time-consuming. Higher speed vs increased
sensitivity.
Genome-size comparative alignment tools











ASSIRC - Accelerated Search for SImilarity Regions in Chromosomes
◦ ftp://ftp.biologie.ens.fr/pub/molbio/ (Vincens et al. 1998)
BLAT –
◦ http://genome.ucsc.edu/cgi-bin/hgBlat?command=start (Kent xxx)
DIALIGN - DIagonal ALIGNment
◦ http://www.gsf.de/biodv/dialign.html (Morgenstern et al. 1998; Morgenstern 1999(
DBA - DNA Block Aligner
◦ http://www.sanger.ac.uk/Software/Wise2/dba.shtml (Jareborg et al. 1999(
GLASS - GLobal Alignment SyStem
◦ http://plover.lcs.mit.edu/ (Batzoglou et al. 2000)
LSH-ALL-PAIRS - Locality -Sensitve Hashing in ALL PAIRS
◦ Email: [email protected] (Buhler 2001)
MegaBlast
◦ http://www.ncbi.nih.gov/blast/ (Zhang 2000)
MUMmer - Maximal Unique Match (mer)
◦ http://www.tigr.org/softlab/ (Delcher et al. 1999)
PIPMaker - Percent Identity Plot MAKER
◦ http://biocse.psu.edu/pipmaker/ (Schwartz et al. 2000)
SSAHA – Sequence Search and Alignment by Hashing Algorithm
◦ http://www.sanger.ac.uk/Software/analysis/SSAHA/
WABA - Wobble Aware Bulk Aligner
◦ http://www.cse.ucsc.edu/~kent/xenoAli/ (Kent & Zahler 2000)
SSAHA
Sequence Search and Alignment by Hashing Algorithm
 Software tool for very fast matching and alignment of DNA
sequences.
 Achieves fast search speed by converting sequence
information into a hash table data structure which can then
be searched very rapidly for matches
 http://www.sanger.ac.uk/Software/analysis/SSAHA/
 Run from the Unix command line
 Need > 1GB RAM (needs a lot of memory)
 SSAHA algorithm best for application requiring exact or
“almost exact” matches between two sequences – e.g. SNP
detection, fast sequence assembly, ordering and orientation of
contigs

Genome Analysis Environment
MAGPIE - Automated Genome Project
Investigation Environment
 PEDANT
 SEALS

Problems with Visualizing Genomes

Alignment programs output often were visualized by text file, which
can be intuitively difficult to interpret when comparing genomes.

Visualization tools needed to handle the complexity and volume of
data and present the information in a comprehensive and
comprehensible manner to a biologist for interpretation.

Genome Alignment Visualization tools need to provide:
◦ interpretable alignments,
◦ gene prediction and database homologies from different sources
◦ Interactive features: real time capabilities, zooming, searching specific
regions of homologies
◦ Represent breaks in synteny
◦ Multiple alignments display
◦ Displaying contigs of unfinished genomes with finished genomes
◦ Handle various data formats
◦ Software availabilty (no black box)
Genome Comparison Visualization Tool

ACT - Artemis Comparison Tool (displays parsed BLAST
alignments; based on Artemis – an annotation tool)
◦ http://www.sanger.ac.uk/Software/ACT/

Alfresco (displays DBA alignments and ...)
◦ http://www.sanger.ac.uk/Software/Alfresco/ (Jareborg & Durbin 2000)

PipMaker (displays BlastZ alignments)
◦ http://bio.cse.psu.edu/pipmaker/ (Schwartz et al. 2000)

Enteric/Menteric/Maj (displays Blastz alignments)
◦ http://glovin.cse.psu.edu/enterix/ (Florea et al. 2000; McClelland et al. 2000)

Intronerator (displays WABA alignments and ...)
◦ http://www.cse.ucsc.edu/~kent/intronerator/ (Kent & Zahler 2000b)

VISTA (Visualization Tool for Alignment) (displays GLASS alignments)
◦ http://www-gsd.lbl.gov/vista/

SynPlot (displays DIALIGN and GLASS alignments)
◦ http://www.sanger.ac.uk/Users/igrg/SynPlot/
Artemis Comparison Tool (ACT)
ACT is a DNA sequence comparison viewer based on
Artemis
- Can read complete EMBL and GenBank entries or
sequence in FASTA or raw format
- Additional sequence feature can be in EMBL, GenBank,
GFF format
- ACT is free software and is distributed under the GNU
Public License
- Java based software
- Latest release 2.0 better support Eukaryotic Genome
Comparison
-
http://www.sanger.ac.uk/Software/ACT/
Salmonella typhi vs. E. coli – SPI-2
G+C
S.typhi
tRNA
phage/IS genes
Pseudogenes
Blast hits
E.coli
Neisseria meningitidis - A vs. B comparison - ACT
A case Study:Comparison of
mouse chromosome 16 and
the human genome:



Mural et al., Science, 2002, 296:1661
Celera group
Synteny with human chr.’s 3,8,12,16,21,22
and rat chr.’s 10,11
Q: Why more breakpoints in mouse-human
than in mouse-rat?
Q: Why more conserved genes in human
than in rat?
•This also can occur between chromosomes
•The longer the divergence time between 2 species, the more
recombination has occurred
•100 million years since human-mouse divergence
•40 million years since rat-mouse divergence
Whole-genome shotgun sequencing:
1. Genome is cut into small sections
2. Each section is hundreds or a few
thousand bp of DNA
3. Each section is sequenced and put
in a database
4. A computer aligns all sequences
together (millions of them from
each chromosome) to form contigs
5. Contigs are arranged (using
markers, etc) to form scaffolds
Q: What are the advantages of this
over the traditional method?
Q: What are the potential sources of
error?
1. Assembly of Mmu16

1.
2.
3.
4.
5.
6.
Total size: 99Mbp
Not one contiguous sequence (contig)
8,635 contigs on 20 “scaffolds”
Average scaffold size: 10Mbp
Number of gaps: 8615
Total size of gaps: ~6Mbp
Total coverage: ~93Mbp
2. Identify genes in Mmu16
1.
2.
3.
4.
5.
Scaffolds of >10kbp were examined (scaffolds larger than 1Mbp were
chopped)
Regions with repeat motifs were ignored using RepeatMasker
Several gene prediction engines use (GenScan, Grail, Fgenes)
Amino acid sequences from open reading frames searched against nr
protein db (NCBI)
Nucleotide searchers (using DNA from across scaffolds) performed
against:
1.
2.
3.
4.
5.
Celera’s gene clusters
Mmu, Rno, & Hsa EST db’s
NCBI’s RefSeq mRNA db
Celera’s dog genomic db
Public pufferfish genomic db
2. Identify genes in Mmu16
6.
7.
8.
9.
1055 genes with high & medium confidence were predicted
Other efforts have identified 1142 genes
After visual annotation inspection, psuedogenes and annotation
errors removed, leaving 731 homologues genes
The genes found were mostly orthologues because they were
reciprocal best matches by BLAST searches.
3. Identify regions of conserved synteny
between Mmu16 and Hsa


1.
2.
Regions of conserved synteny predicted by sequence similarity and by
protein comparisons
Synteny based on sequence comparisons:
Syntenic anchors were located - regions with high (80%) similarity over
short distances (~200bp or more).
Average distance between anchors is 8kbp, but there are gaps as large as
707kbp in the mouse and 3.4Mbp in the human
3. Identify regions of conserved synteny
between Mmu16 and Hsa
3.
4.
5.
56% of anchors were in mouse genes - exons mostly
44% in intergenic regions
Relatively density is independent of coding/noncoding - making the anchors
an important marker of synteny (in addition to genes)
Human chr.
16
8
12
22
3q27-29
3q11.1-13.3
21
Mmu len.
10,461
1,284
363
2,081
13,557
41,660
22,327
Hsa len.
12,329
1,491
306
2,273
16,461
46,493
28,421
No. anchors
1,429
121
31
418
1,714
5,485
2,127
bad anch. (% incon.)
21 (1.5)
1 (0.8)
3 (9.7)
8 (1.9)
18 (1.0)
63 (1.1)
27 (1.3)
Orthologues
87
6
3
30
107
165
111
Summary
Performance evaluation of gene-finding programs
 Comparative genomics
 Comparative genomics analysis example

Acknowledgement

Chuong Huynh (NIH)