1. Functional and structural genomics using PEDANT
陽明生技所
生物資訊學程
林千涵

2. Introduction
With increasing biological sequence data, it need a system with ability of storing and retreving tens of gigabytes of data, a mature database management system, and a good visualization tools
From case-oriented sequence analysis work to automated large-scale genome annotation

3. Introduction-PEDANT Difference of existing genome analysis programs
protein oriented vs. DNA oriented analysis
interactive work vs. commandline operation
bioinformatics method applied
user interface
conveniency feature, project management and data editors
fidelity of result produced
Benchmark may vary in terms of chosen of balance between sensitivity and selectivity of the analyses
PEDANT (Protein Extraction, Description, and ANalysis Tool) was available in mid-1997(use FASTA as similarity search)
a workhorse for general bioinformatics research
a common framework for a number of genome analysis projects
a complete database of automated genomes
a tool for routine analysis of large amounts of genomic contigs and ESTs

4. System Architecture Overview
database module: storing, modifying and accessing data
processing module: bioinformatics computations
user interface: web based communication

5. System Architecture-Cont.
Data access
primary table: store raw data (ex DNA, protein sequences and program results ex BLAST output )
secondary table: parsed program results
simplified schema
Operation in command line mode
applying bioinformatics methods to sequences
parsing data tables
querying the resulting databases
Web interface
No static HTML pages required
DNA and Protein viewers make direct access to the SQL tables
Implementation and system requirements
Perl 5, and C++ for graphical viewer
Performance
parallel capabilities

6. Schema

7. Bioinformatics Method
Overview of the PEDANT processing pipeline
identification of coding regions and various analysis genetics elements
homology search
detection of protein motifs, prediction of secondary structure and other protein features and sensitive fold recognition
automatically attributed to pre-defined functional categories
Prediction of genes and other genetic elements
Table 1
choose one of 15 genetic codes
Functional and structural categories
similarity search : PSI-BLAST(Position-Specific Iterated BLAST)
special datasets: MIPS, COG, PROSITE, PFAM and BLOCKS
significant matches of PIR: annotations, keywords, enzyme classification and superfamily information
with significant relationship of PDB, secondary structure information: STRIDE(upper case), PREDATOR(lower case)
low complexity region, membrance regions, coiled coils and signal peptides
comparison of SCOP with IMPALA
functional
structural

8. Table 1

9. Bioinformatics Method-Cont.
Yeast biological role categories
first system of biological role of categories : E.Coli
MIPS: advanced hierarchical functional catalogue (Yeast)
Multidimensionality-protein:gene is M:M
automated assignment to MIPS is first approximation, will be refined by manual annotation
Distribution of ORFs
Visualization
a integrated, hypertext-linked protein report with calculated parameters and sequences as reference for further manual annotation
Protein report page

10. Distribution of ORFs

11. Protein report page

12. Bioinformatics Method-Cont.2
Automatic versus manual annotation
Problem of error propagation
erroneous annotation by human error and spurious similarity hits
with filtering algorithms and domain structure ?
quality improvement of manual review of human experts !
Manual annotation
Catalogue independent
Flexibility: first place in higher category and later step move to the finer categories
528 categories: 20 main categories and 6 levels
confidence levels: “reject”, “low”, “medium”, “high” and default is “auto”
Data release management
new release data can be intelligently merged with existing data pool
transfer manual annotation between subsequent data release
“manual” field: “yes” or ”no” and default is “no” initially
example: a PFAM domain identified in new release ORF is “manual: no” and “conf: auto”

13. Manual annotation transfer
Two genes fuse to one contig
Two contigs fuse to one
Gene boundary change
Appears new gene

14. The PEDANT Genome Database
Annotation of publicly available completely sequenced and unfinished genomes
Genome annotated by MIPS
Completely sequenced and published genomic sequences
Unfinished and/or unpublished genomics sequences
gene prediction by ORPHEUS, allow large overlaps between ORFs
PEDANT as a structural genomics resource-0.3M proteins
class-based approach, cost-saving
(i)non-redundant protein sequence databases
(ii)PSI-BLAST search with SCOP against (I) abd saving resulting profiles
(iii)construct a SCOP profile library using IMPALA
(iv)IMPALA search with each genomic sequence against SCOP library
same procedure for nr PDB sequence database
performance of IMPALA
Cross-genome comparison
treat each genome as an individual contig : creat cross-genome datasets without any modification
44 genomes

15. Performance of IMPALA

16. Applications Arabidopsis thaliana chromosome IV
3744 predicted protein coding genes
roughly 30% are known proteins or strongly similar to known proteins
multi-cellular organisms has higher all-alpha and smaller mixed alpha/beta structural domains ratio to unicellular species
Assembled human transcripts
human UniGene subjected PEDANT analysis, compare over contigs
this MySQL DB is close to 8GB
acceptable query time show the suitability of PEDANT for large-scale EST sequencing projects
Analysis of the GroEL substrates
GroEL: a common E.Coli chaperonin
structural motif common in 52 substrates relying on GroEL for folding in vivo : two or more alpha/beta domains involving buried beta-sheets with large hydrophobic surfaces--easy aggregation

17. Classification of predicted genes
Classification by the degree of homology to functionally characterized proteins based on BLAST scores

18. Summary and Outlook
PEDANT is a useful tool for genome annotation and bioinformatics research
It can automated and manual assignment of gene product to functional and structural categories
extensive hyperlinked protein report and advanced viewers
Outlook
better decision rules need to be employed
manually annotate predicted genetics eelments(ex. LTRs)
supporting Oracle RDBMS
automatic gene prediction pipeline for higher eukaryotes
interactive capabilities