1. Bioinformatics and Genome Annotation
Shane C Burgess

2. NIH WORKING DEFINITION OF BIOINFORMATICS AND COMPUTATIONAL BIOLOGY
July 17, 2000
Bioinformatics: Research, development, or application of computational tools and approaches for expanding the use of biological, medical, behavioral or health data, including those to acquire, store, organize, archive, analyze, or visualize such data.
Computational Biology: The development and application of data-analytical and theoretical methods, mathematical modeling and computational simulation techniques to the study of biological, behavioral, and social systems.

3. Biocomputing:computational biology & bioinformatics
Gene Ontology Consortium members 3

4. Dr Fiona McCarthy
Dr Susan Bridges
Dr Teresia Buza
Dr Nan Wang
Cathy Grisham
Dr Divya Pedinti
Philippe Chouvarine
Lakshmi Pillai

5. Sequencing is getting cheaper
Cost of human or similar sized genome
Source: Richard Gibbs,
Baylor College of Medicine
and biocomputing becomes more of an issue.

6. Complexity
Sequence itself and from all it’s compatriots and assorted microbes
SNPs
Transcripts (all of them…don’t forget alternative splicing, starts)
CNVs
Epigenetic changes to DNA
Proteome (expression, epigenetics, PTMs, location, flux, enzyme kinetics)
Metabolites
Phenotypes
Drugs
B. Statistical. 1. Multiple testing problem. 2. Search space
Both have potential computationally-intensive solutions (Monte Carlo/Resampling/ Permutation/Bootstrap and target/decoy).
C. Information: publications are no longer the sole source of “valid” or “legitimate” information.
Trusted databases and not just publications used as research sources; not just data but also community annotations etc
D. Biocomputing issues: LOCAL--storage, compute power (CPUs days), RAM; DISTANT– linking, data movement, cyberinfrastucture (hard, soft and human).
E. How and who?

7. Titus Brown, Mich. SU

8. Storage costs
A. Simple Storage Service (S3) e.g. Amazon. For the first 50 TB = 15 US cents/Gb ($7,500/50 TB) plus pay for data transfer and operations. VS
Buy, store and scale as needed e.g. Web Object Scaler (WOS)
Immediate or “longer” term solution
Putting Genomes in the Cloud. Making data sharing faster, easier and more scalable.
By M. May, May 18, 2010.

9. 10 Gigabits (Gb)/second

10. Annotation: Nomenclature, Structural & Functional
Structural Annotation:
Open reading frames (ORFs) predicted during genome assembly
predicted ORFs require experimental confirmation
Functional Annotation:
annotation of gene products = Gene Ontology (GO) annotation
initially, predicted ORFs have no functional literature and GO annotation relies on computational methods (rapid)
functional literature exists for many genes/proteins prior to genome sequencing
Gene Ontology annotation does not rely on a completed genome sequence

11. Chicken Gene Nomenclature
Livestock Gene Nomenclature:
Jim Reecy et al., International Society for Animal Genetics from 26th – 30th July 2010, Edinburgh
Chicken Gene Nomenclature
1995: chicken gene nomenclature will follow HGNC guidelines
2007: chicken biocurators begin assigning standardized nomenclature
2008: first CGNC report; NCBI begins using standardized nomenclature & CGNC links
2010: first dedicated chicken gene nomenclature biocurator; NCBI/AgBase/Marcia Miller – structural annotation & nomenclature for MHC regions (chr 16)
Chicken gene nomenclature database – UK & US databases sharing and co-coordinating data.

12.
Available via BirdBase & AgBase

13. Experimental Structural genome annotation Proteogenomic mapping

14. Problems with Current Structural Annotation Methods
EST evidence is biased for the ends of the genes
Computational gene finding programs
Misidentify some, and especially short, genes, genes.
Overlook exons
Incorrectly demarcate gene boundaries, especially splice junctions
Annotation of the open reading frames (ORFs) (i.e., gene prediction) in genomes is most often performed computationally based on features in the nucleic acid sequence.
Homology searches method

15. Proteogenomic Mapping
Combines genomic and proteomic data for structural annotation of genomes
First reported by Jaffe et al. at Harvard in 2004 in bacteria
McCarthy et al first applied in chicken (one of the first uses in a eukaryote; the other two in human).
Improves genome structural annotation based on expressed protein evidence
Confirms existence of predicted protein-coding gene
Identifies exons missed by gene finder
Corrects incorrect boundaries of previously identified genes
Identifies new genes that the gene finding programs missed

16. CCV genome was sequenced in 1992
But only 12 of predicted 76 ORFs confirmed to exist as proteins.
Confirmed 37/76.
Identified 17 novel ORFs that were not predicted.

18. Structural Annotation of the Chicken Genome
Location of genes on the genome
Computational gene finding programs such as Gnomen (NCBI) based on Markov Models and also use
ESTs
Known proteins
Sequence conservation

19. ePST Generation Process
Peptide nucleotide sequence
chromosome
Map peptide nucleotide sequence to chromosome

20. Search against protein Database
Biological Sample
Trypsin Digestion
LC ESI-MS/MS Data
Search against protein Database
Search against genome translated in 6 reading frame
Peptide matches
Peptide matches
Generate ePST (expressed PeptideSequence Tags) from peptides matching genome only
Confirm predicted protein-coding gene
Correction / validation of genome annotation
Novel protein-coding gene

21. ePST Generation Process
Peptide nucleotide sequence
Stop codon
chromosome
Locate first downstream in-frame stop codon or canonical splice junction

22. ePST Generation Process
Peptide nucleotide sequence
Stop codon
chromosome
Locate upstream canonical splice junction or in-frame stop

23. ePST Generation Process
Peptide nucleotide sequence
Stop codon
chromosome
Start codon
Find 1st start codon between in-frame stop and peptide

24. ePST Generation Process
chromosome
Use splice junction or in-frame start as beginning of ePST

25. ePST Generation Process
chromosome
ePST coding nucleotide sequence
Translate
Expressed Peptide Sequence Tag (ePST) amino acid sequence

30. Functional annotation

31. 2 4 6 8 10 12 14 16 18 70 75 80 85 90 95 00 05
No. x 106
No.
5000
10000
15000
20000
25000
‘00
‘01
‘02
‘03
‘04
‘05
‘06
‘07
‘08
‘09
YEAR

32. Functional Understanding
Canonical and other Networks
Ontologies
Functional Understanding
GO Cellular Component
GO Biological Process
GO Molecular Function
BRENDA
Pathway Studio 5.0
Ingenuity Pathway Analyses
Cytoscape
Interactome Databases

33. Biological interpretation
Gene Ontology
Network Modeling
Derived
Implied
Physiology
(= Cellular Component + Biological Process + Molecular Function)

34. What is the Gene Ontology?
“a controlled vocabulary that can be applied to all organisms even as knowledge of gene and protein roles in cells is accumulating and changing”
the de facto standard for functional annotation
assign functions to gene products at different levels, depending on how much is known about a gene product
is used for a diverse range of species
structured to be queried at different levels, eg:
find all the chicken gene products in the genome that are involved in signal transduction
zoom in on all the receptor tyrosine kinases
human readable GO function has a digital tag to allow computational analysis of large datasets
COMPUTATIONALLY AMENABLE ENCYCLOPEDIA OF GENE FUNCTIONS AND THEIR RELATIONSHIPS

35. GO is the “encyclopedia” of gene functions captured, coded and put into a directed acyclic graph (DAG) structure.
In other words, by collecting all of the known data about gene product biological processes, molecular functions and cell locations, GO has become the master “cheat-sheet” for our total knowledge of the genetic basis of phenotype.
Because every GO annotation term has a unique digital code,
we can use computers to mine the GO DAGs for granular functional information.
Instead of having to plough through thousands of papers at the library and make notes and then decide what the differential gene expression from your microarray experiment means as a net affect, the aim is for GO to have all the biological information captured and then retrieve it and compile it with your quantitative gene product expression data and provide a net affect.

36. Use GO for…….
Determining which classes of gene products are over-represented or under-represented.
Grouping gene products.
Relating a protein’s location to its function.
Focusing on particular biological pathways and functions (hypothesis-testing).

38. Many people use “GO Slims” which capture only high-level terms which are more often then not extremely poorly informative and not suitable for hypothesis-testing.
“GO Slim”
In contrast, we need to use the deep granular information rich data suitable for hypothesis-testing

39. Sourcing displaying GO annotations: secondary and tertiary sources.

41. GO Consortium: Reference Genome Project
Limited resources to GO annotate gene products for every genome
rely on computational GO annotations
most robust method is to transfer GO between orthologs
Reference genome project: goal is to produce a “gold standard” manually biocurated GO annotation dataset for orthologous genes
12 reference genomes – chicken is only agricultural species
Chicken RGP contributions provided via USDA CSREES MISV

42. RGP & Taxonomy checks
Transferring GO annotation between orthologs requires:
determining orthologs – computational prediction followed by manual curation
developing ‘sanity’ checks to ensure transferred functions make sense phylogenetically (eg. no lactating chickens!)

43. Further taxon checking comments may be added here, or contact the AgBase database.

44. AgBase Quality Checks & Releases
AgBase Biocurators
‘sanity’ check
AgBase
biocuration
interface
‘sanity’ check
& GOC QC
AgBase database
GO analysis tools
Microarray developers
‘sanity’ check
UniProt db
QuickGO browser
GO analysis tools
Microarray developers
EBI GOA Project
‘sanity’ check: checks to ensure all appropriate information is captured, no obsolete GO:IDs are used, etc.
‘sanity’ check
& GOC QC
Public databases
AmiGO browser
GO analysis tools
Microarray developers
GO Consortium
database

45. Gene Products annotated
Comparing AgBase & EBI-GOA Annotations
14,000
computational
manual - sequence
12,000
manual - literature
10,000
Gene Products annotated
8,000
Complementary to EBI-GOA: Genbank proteins not represented in UniProt & EST sequences on arrays
6,000
4,000
2,000
AgBase
EBI-GOA
AgBase
EBI-GOA
Chick
Chick
Cow
Cow
Project

46. Contribution to GO Literature Biocuration
AgBase
EBI GOA
Chicken
97.82%
EBI-IntAct
Roslin
HGNC
< 0.50%
UCL-Heart project
MGI
Cow
Reactome
88.78%
< 1.50%

47. INPUT:
functional genomics data (e.g. Microarray data)
GOanna
Biocuration from literature
Manual interpretation of GOanna output
gene products with NO orthologs OR with orthologs but NO GO annotations
GOModeler
Generic: qualitative data presentation. Analysis can only be changed if user has programming skills
Specific: user-defined, hypothesis-driven, quantitative data presentation
must wait on experimental evidence or new electronic inference
NO literature or specialist knowledge that can be used to make GO annotations
gene products with orthologs and GO annotations
gene products with NO GO annotations
gene products with GO annotations
BLAST output
biocurated annotations from literature or specialist knowledge
GOSlimViewer
GORetriever
data visualization
ArrayIDer
GOanna2ga
comprehensive GO annotation
(existing GO analysis programs)
GA2GEO
GAQ Score
Figure 4: GO Tools available at the AgBase website. We have added tools to help users get comprehensive GO annotations for functional modeling (indicated by blue boxes) and extended the ID types that existing tools (green boxes) accept. The inset shows the types of IDs now accepted.

48. To request a workshop contact
2010 GO Training Opportunities
- on site training by request/interest
- webinar: notification via ANGENMAP & GO discussion groups
To request a workshop contact
Fiona McCarthy OR

49. GO training 2009 Workshop hosts: ISU – Dr Susan Lamont
Workshop Surveys 10 20 30 40 50 60
Topics covered were relevant
Topics were well explained
I am confident in using GO for modeling
I am confident I can get GO questions answered
I would recommend this workshop
% of respondents
strongly agree
agree
uncertain
disagree
strongly disagree
GO training 50
100
150
200
2007
2008
2009
Year workshops offered
No. of people
Annual
Cumulative
2009 Workshop hosts:
ISU – Dr Susan Lamont
NCSU – Dr Hsiao-Ching Liu

51. Number of participants: 25 Number of arrays: 22 Number of votes: 41
Chicken Array Usage
Number of participants: 25
Number of arrays: 22
Number of votes: 41
Bovine array usage
Number of participants: 26
Number of arrays: 26
Number of votes: 42
ARK-Genomics
Affymetrix
Agilent 44K array
UD 7.4K Metabolic/Somatic
UD_Liver_3.2K
Arizona 20.7K
Neuroendocrine
UIUC 13.2K
Agilent 44k
Bovine Total Leukocyte cDNA
Affymetrix
UIUC 7,872-element

52. Quality improvement Microarray annotations

53. Most microarray analysis tools do not readily accept EST clone names (abundantly on arrays).
Manual re-annotation of microarrays is impracticable
Retrieves the most recent accession mapping files from public databases based on EST clone names or accessions and rapidly generates database accessions.
Fred Hutchinson Cancer Research Centre 13K chicken cDNA array
structurally re-annotated 55% of the array; decreased non-chicken functional annotations by 2 fold; identified 290 pseudogenes, 66 of which were previously incorrectly annotated.

55. Zhou H, Lamont SJ: Global gene expression profile after Salmonella enterica Serovar enteritidis challenge in two F8 advanced intercross chicken lines. Cytogenet Genome Res 2007;117: (DOI: / )

59. Increased the pathway coverage of several major immune response pathways and provided more comprehensive modelling of signalling pathways e.g. FAS :originally not annotated but now pathways involving FAS identified.
Confirm and consolidate previous suggestions that CD3e, IL-1β, and CCL5 differential expression involved in the immune response to SE. Chicken-specific functional annotation of these genes allowed identification of these gene’s related pathways with statistical confidence.
Identified additional genes involved in major immune pathways important in bacterial gut disease but not identified in the original work e.g. tyrosine phosphatase type IVA member 1 (PTP4A1); CD28; T-cell co-stimulator (ICOS, CD287) and NK-lysin and associated pathway genes.

60. Bacterial functional genomic responses to structural differences in explosive compounds.
KTR9 and
V. fischeri proteomics

61. Quantifying re-annotation
Metrics
Granularity Specificity
# previous annotations # chicken annotations
# re-annotations # human/mouse annotations
Quality
Gene Ontology Annotation Quality (GAQ) score

62. Reads in annotated gene regions + 20 kb radius
Reads in “RNAFAR” regions i.e. clustered reads forming novel transcripts (these reads do not belong to any gene model the reference set and can either be assigned to neighboring gene models, if they are within a specified threshold radius, or assigned their own predicted transcript model.
Repeats with > 10 alignments
Reads overlapping annotated repeat regions
Unmapped reads
Other (regulatory, etc. do not include reads discarded as poor quality).
DoD: Bobwhite Quail Toxicogenomics
Mean GAQ score 62

64. GO Cellular Component DAG

65. Differential Detergent Fractionation
DDF Fraction 1 2 3 4
2007. Non-electrophoretic differential detergent fractionation proteomics using frozen whole organs. Rapid Commun Mass Spectrom 21:
2007. Sequential detergent extraction prior to mass spectrometry analysis. Methods in Molecular Medicine: Proteomic analysis of membrane proteins. Humana Press. 117 (1-4):
2005. Differential detergent fractionation for non-electrophoretic eukaryote cell proteomics. Journal of Proteome Research. 4 (2),

66. M C N Sub-cellular localization of pro-PCD proteins. B-cells Stroma
One mechanism controlling PCD is the release of “pro-death” proteins mitochondria into the cytoplasm or nucleus.
B-cells
Stroma C
CytC
Apaf1
AMID
EndoG
AIF
Smac M N

67. Neoplastic compared to Hyperplastic
100000
mRNA
10000
1000
100 10 1
IL-2
IL-4
IL-6
IL-8
IL-10
IL-12
IL-13
Neoplastic compared to Hyperplastic
lymphoma cells (%)
IL-18
IFNg
TGFb
CTLA-4
GPR-83
SMAD-7 4
Protein 3
IL-10 2
IL-12 1
IL-6
IL-13
IL-8
IL-2
IL-4
IL-18
TGFb
CTLA-4
GPR-83 -1
IFNg
SMAD-7 -2 -3
Cancer Immunology and Immunotherapy, : 67

68. IL-18 distribution: it matters where proteins are2 3 4
IL-18 distribution: it matters where proteins are 1 2 3 4
DDF Fraction
Neoplastic Lymphocytes
(T-reg)
Hyperplastic Lymphocytes
Extracellular
Nuclear 15 20 25 30 35 5 10 10 20 30 40 50 60 70 80
Shack et al., Cancer Immunology and Immunotherapy, : 68

70. Pig Bindu Nanduri Translation to clinical research
Total mRNA and protein expression was measured from quadruplicate samples of control, electroscalple and harmonic scalple-treated tissue.
Differentially-expressed mRNA’s and proteins identified using Monte-Carlo resampling1.
Using network and pathway analysis as well as Gene Ontology-based hypothesis testing, differences in specific phyisological processes between electroscalple and harmonic scalple-treated tissue were quantified and reported as net effects.
(1) Nanduri, B., P. Shah, M. Ramkumar, E. A. Allen, E. Swaitlo, S. C. Burgess*, and M. L. Lawrence* Quantitative analysis of Streptococcus Pneumoniae TIGR4 response to in vitro iron restriction by 2-D LC ESI MS/MS. Proteomics 8,

72. Proportional distribution of protein functions differentially-expressed by Electro and Harmonic Scalpel
HYPOTHESIS TERMS
Total differentially-expressed proteins: 433
Harmonic Scalpel
Total differentially-expressed proteins: 509
Electroscalpel
immunity (primarily innate)
inflammation
Wound Healing
Lipid metabolism
response to Thermal Injury
angiogenesis
hemorrhage

73. Net functional distribution of differentially-expressed proteins
Harmonic Scalpel
Electroscalpel
hemorrhage
sensory response to pain
angiogenesis
response to thermal injury
Lipid metabolism
Wound healing
classical inflammation
(heat, redness, swelling, pain, loss of function)
immunity (primarily innate) 8 6 4 2 2 4 6
Relative bias