Bioinformatics and cancer

1
Bioinformatics and cancer

• How an experimental science can benefit from
  information technology

2
What is bioinformatics?

• Application of computer technology to biological
  problems
• Extraction of biological information from raw
  data
• Presentation of complex results in an
  understandable form
• Generation of testable hypotheses for the
  experimentalists

3
How is bioinformatics relevant to cancer?

• Record keeping, particularly in complex,
  multi-center studies
• Immunoscope
• Compiling and integrating information derived
  from multiple sources
• SEREX
• Leveraging the information derived from genome
  and transcriptome sequencing projects
• ORESTES

4
The SEREX database

• Single repository for all SEREX data
• Unified format for storing and accessing data
• Unified set of tools for first-level analysis of
  new sequences
• Identification of recurrent epitopes
• Curation and annotation of data

5
Primary SEREX data

• Identity of clone
• Library, origin of serum, clone identifiers
• Sequence of clone
• Nucleotide, protein if unambiguous
• Experimental data
• Reactivity with autologous and hetereologous sera
• Tissue-specific expression

6
Derived SEREX data

• Identity of gene from which cDNA was derived
• Full sequence of the cDNA
• Sequence of the corresponding protein
• Similarities in the databases
• Chromosomal localization

7
Nucleotide sequence analysis methods

• Compare to SEREX database
• Find if epitope already identified by others
• Compare to human section of EMBL
• Find if sequence is derived from known gene
• Compare to Unigene database
• Identify gene or EST cluster from which sequence
  is derived
• Get information about chromosomal localization,
  specificity of expression
• Compare to protein databases
• Get information about candidate ORFs, taking
  frameshifts into account

8
Protein sequence analysis methods

• Compare to protein databases
• Try to identify potential homologs
• Compare to profile databases
• Identify sequence motifs indicative of structure
  and/or function
• Search for HLA-binding peptides
• Intrinsic methods
• Look for coiled-coils, transmembrane segments,
  signal peptides, etc.

9
Current status of database

• 1806 entries (Sep. 2000)
• Reactivity 359 testis, 206 stomach, 310 breast,
  157 kidney, 102 colon, 37 lung, 73 melanoma, 180
  cell lines
• 1480 entries in the public part of the database
• 593 sequences match at least one other sequence
• Many genes are novel

10
Re-engineering the SEREX database

• Convert to true relational database
• Standardize the annotation
• Link to external databases
• SWISS-PROT, GeneCards, LocusLink, RefSeq
• Integrate into Cancer Immunome Database

11
Gene discovery tools

• General goal convert raw data to a form where it
  can be searched efficiently and successfully
• Tools developed at the SIB
• EST clustering and assembly software
• Genome contig reconstitution
• Program to find and correct coding regions in
  low-quality sequences

12
ESTScan, TrEST and TrGEN

• ESTScan is a tool that uses a special hidden
  Markov model of coding regions to find and
  correct coding sequences in ESTs
• TrEST is a virtual protein database derived from
  EST contigs using ESTScan
• TrGEN is a virtual protein database derived from
  raw genome sequences using GENSCAN
• Both are very rich sources of novel genes

13
Hunting for homologs of NY-ESO-1

• NY-ESO-1 is a prototype CT antigen, to which
  cancer patients mount both humoral and cellular
  responses
• There is a second human gene, LAGE-1, also
  located on chr X and highly similar to NY-ESO-1
• Are there other human genes in the same family?
• Are there homologs in other species?

14
Methodology

• Identify similar regions in NY-ESO-1 and LAGE-1,
  and make profile
• Use profile to search TrEST and TrGEN, plus
  traditional protein databases (SWISS-PROT,
  TrEMBL)
• Use profile to search EST contigs and predicted
  genomic exons in framesearch tolerant mode
• Use hits from profile search to refine profile
  and reiterate database searches

15
Human homologs

• First search identifies a predicted CDS and a
  Unigene cluster mapping to Xp28
• The corresponding mRNA and protein are in the
  databases (ITBA2), but translated in the wrong
  frame!
• Second search identifies a new Unigene cluster,
  whose closest current homolog is a pseudogene on
  chr9
• Therefore, there are at least four human family
  members

16
Rodent homologs

• Still very little genomic sequence, so have to
  rely on ESTs alone
• In mouse, pick up two clusters that are similar
  to each other and to ITBA2
• In rat, pick up only one cluster similar to the
  mouse ones
• One mouse mRNA checked for tissue specificity
  turned out not to be CT antigen

17
More distant homologs

• For genes centrally involved in development or
  metabolism, expect homologs in most eukaryotes
• Found ESO-1 homologs in several fully sequenced
  genomes Drosophila, C. elegans, S. pombe (but
  not S. cerevisiae)
• In Drosophila, protein was misannotated as
  ribosomal, because is it located adjacent to L34
  protein from 60S subunit

18
Alignment of homologous regions
19
Lessons learned

• EST and unannotated genome data are still a rich
  source of information for gene discovery
• Much of the existing annotation is erroneous,
  even if it was not done automatically
• Bioinformatics approaches can suggest new
  experimental avenues

20
The ORESTES project

• Sponsors Ludwig Institute for Cancer Research,
  FAPESP (São Paulo State funding agency)
• Goal to obtain EST sequences from the
  under-represented, often coding, central portions
  of mRNAs
• Methodology use low-stringency semi-random
  priming followed by PCR, producing low complexity
  libraries
• Results over 500000 ESTs produced, of which
  half produce novel information

21
The Human Transcriptome project

• Sponsors NCI, LICR, FAPESP
• Goal to provide a comprehensive and
  experimentally validated catalog of human
  transcripts
• Methodology create a stable index of transcripts
  by using poly(A) tags, extend this by a
  combination of expermental and bioinformatic
  methods

22
The EUCIP project

• Sponsors European Union, LICR
• Goal to examine in detail the cancer biology of
  genes identified using the SEREX method
• Methodology for each gene, examine
  cancer-related alterations, patterns of
  expression, tissue distribution and immune
  responses document in integrated database

23
The DNA chip consortium

• Partners LICR, ICRF, Sanger Centre
• Goal to produce cDNA chips representing the
  entire human transcriptome
• Current state two 5000-feature chips produced
  routinely, expect non-redundant 10k chips at
  years end

24
Credits

• Philipp Bucher (ISREC)
• Christian Iseli (LICR)
• Brian Stevenson (LICR)
• Dmitry Kuznetsov (LICR)
• Andy Simpson and Sandro de Souza (LICR São Paulo)