Medical Informatics

1
Medical Informatics
Bioinformatics
2
Mining Bio-Medical Mountains How Computer Science
can help Biomedical Research and Health Sciences
Anil Jegga Division of Biomedical Informatics,
Cincinnati Childrens Hospital Medical Center
(CCHMC) Department of Pediatrics, University of
Cincinnati http//anil.cchmc.org Anil.Jegga_at_cchmc.
org
3
Acknowledgement

• Biomedical Engineering/Bioinformatics
• Jing Chen
• Sivakumar Gowrisankar
• Vivek Kaimal
• Computer Science
• Amit Sinha
• Mrunal Deshmukh
• Divya Sardana
• Electrical Engineering
• Nishanth Vepachedu

4
Two Separate Worlds..
Medical Informatics
Bioinformatics the omes
PubMed
Proteome
Disease Database
Patient Records
OMIM Clinical Synopsis
Clinical Trials
382 omes so far and there is UNKNOME too -
genes with no function known http//omics.org/inde
x.php/Alphabetically_ordered_list_of_omics
With Some Data Exchange
5
now. The number 1 FAQ
How much biology should I know??
No simple or straight-forward answer
unfortunately!
But the mantra is Interact routinely with
biologists OR Work with the biologists or the
biological data
6
But I want to learn some basics

• http//www.ncbi.nlm.nih.gov/Education
• http//www.ebi.ac.uk/2can/
• http//www.genome.gov/Education/
• http//genomics.energy.gov/
• Books
• Introduction to Bioinformatics by Teresa Attwood,
  David Parry-Smith
• A Primer of Genome Science by Gibson G and Muse
  SV
• Bioinformatics A Practical Guide to the Analysis
  of Genes and Proteins, Second Edition by Andreas
  D. Baxevanis, B. F. Francis Ouellette
• Algorithms on Strings, Trees, and Sequences
  Computer Science and Computational Biology by Dan
  Gusfield
• Bioinformatics Sequence and Genome Analysis by
  David W. Mount
• Discovering Genomics, Proteomics, and
  Bioinformatics by A. Malcolm Campbell and Laurie
  J. Heyer

7
And the other FAQs.

• What bioinformatics topics are closest to
  computer science?
• Should computer science departments involve
  themselves in preparing their graduates for
  careers in bioinformatics?
• And if so, what topics should they cover?
• And how much biology should they be taught?
• Lastly, how much effort should be expended in
  re-directing computer scientists to do work in
  bioinformatics?

Cohen, 2005 Communications of the ACM
8
Issues to be considered..

• Computer science Vs molecular biology Subject
  Scientists - Cultural differences
• Current goals of molecular biology, genomics (or
  biomedical research in a broader sense)
• Data types used in bioinformatics or genomics
• Areas within computer science of interest to
  biologists
• Bioinformatics research - Employment
  opportunities

9
Biological Challenges - Computer Engineers

• Post-genomic Era and the goal of bio-medicine
• to develop a quantitative understanding of how
  living things are built from the genome that
  encodes them.
• Deciphering the genome code
• Identifying unknown genes and assigning function
  by computational analysis of genomic sequence
• Identifying the regulatory mechanisms
• Identifying their role in normal
  development/states vs disease states

10
Biological Challenges - Computer Engineers

• Data Deluge exponential growth of data silos and
  different data types
• Human-computer interaction specialists need to
  work closely with academic and clinical
  biomedical researchers to not only manage the
  data deluge but to convert information into
  knowledge.
• Biological data is very complex and interlinked!
• Creating information systems that allow
  biologists to seamlessly follow these links
  without getting lost in a sea of information - a
  huge opportunity for computer scientists.

11
Biological Challenges - Computer Engineers
A major goal in molecular biology is Functional
Genomics Study of the relationships among genes
in DNA their function in normal and disease
states

• Networks, networks, and networks!
• Each gene in the genome is not an independent
  entity. Multiple genes interact to perform a
  specific function.
• Environmental influences Genotype-environment
  interaction
• Integrating genomic and biochemical data together
  into quantitative and predictive models of
  biochemistry and physiology
• Computer scientists, mathematicians, and
  statisticians will ALL be an integral and
  critical part of this effort.

12
Informatics Biologists Expectations

• Representation, Organization, Manipulation,
  Distribution, Maintenance, and Use of
  information, particularly in digital form.
• Functional aspect of bioinformatics
  Representation, Storage, and Distribution of
  data.
• Intelligent design of data formats and databases
• Creation of tools to query those databases
• Development of user interfaces or visualizations
  that bring together different tools to allow the
  user to ask complex questions or put forth
  testable hypotheses.

13
Informatics Biologists Expectations

• Developing analytical tools to discover knowledge
  in data
• Levels at biological information is used
• comparing sequences predict function of a newly
  discovered gene
• breaking down known 3D protein structures into
  bits to find patterns that can help predict how
  the protein folds
• modeling how proteins and metabolites in a cell
  work together to make the cell function.

14
Finally.What does informatics mean to
biologists?

• The ultimate goal of analytical bioinformaticians
  is to develop predictive methods that allow
  biomedical researchers and scientists to model
  the function and phenotype of an organism based
  only on its genomic sequence. This is a grand
  goal, and one that will be approached only in
  small steps, by many scientists from different
  but allied disciplines working cohesively.

15
Biology Data Structures

• Four broad categories
• Strings To represent DNA, RNA, amino acid
  sequences of proteins
• Trees To represent the evolution of various
  organisms (Taxonomy) or structured knowledge
  (Ontologies)
• Sets of 3D points and their linkages To
  represent protein structures
• Graphs To represent metabolic, regulatory, and
  signaling networks or pathways

16
Biology Data Structures

• Biologists are also interested in
• Substrings
• Subtrees
• Subsets of points and linkages, and
• Subgraphs.

Beware Biological data is often characterized by
huge size, the presence of laboratory errors
(noise), duplication, and sometimes unreliability.
17
Support Complex Queries A typical demand

• Get me all genes involved in or associated with
  brain development that are differentially
  expressed in the Central Nervous System.
• Get me all genes involved in brain development in
  human and mouse that also show iron ion binding
  activity.
• For this set of genes, what aspects of function
  and/or cellular localization do they share?
• For this set of genes, what mutations are
  reported to cause pathological conditions?

18
Model Organism Databases Common Issues

• Heterogeneous Data Sets - Data Integration
• From Genotype to Phenotype
• Experimental and Consensus Views
• Incorporation of Large Datasets
• Whole genome annotation pipelines
• Large scale mutagenesis/variation projects
  (dbSNP)
• Computational vs. Literature-based Data
  Collection and Evaluation (MedLine)
• Data Mining
• extraction of new knowledge
• testable hypotheses (Hypothesis Generation)

19
Bioinformatic Data-1978 to present

• DNA sequence
• Gene expression
• Protein expression
• Protein Structure
• Genome mapping
• SNPs Mutations

• Metabolic networks
• Regulatory networks
• Trait mapping
• Gene function analysis
• Scientific literature
• and others..

20
Human Genome Project Data Deluge
No. of Human Gene Records currently in NCBI
29413 (excluding pseudogenes, mitochondrial genes
and obsolete records). Includes 460 microRNAs
NCBI Human Genome Statistics as on February12,
2008
21
The Gene Expression Data Deluge
Till 2000 413 papers on microarray!
Problems Deluge! Allison DB, Cui X, Page GP,
Sabripour M. 2006. Microarray data analysis from
disarray to consolidation and consensus. Nat Rev
Genet. 7(1) 55-65.
22
Information Deluge..

• 3 scientific journals in 1750
• Now - gt120,000 scientific journals!
• gt500,000 medical articles/year
• gt4,000,000 scientific articles/year
• gt16 million abstracts in PubMed derived from
  gt32,500 journals

23
Data-driven Problems..

• Generally, the names refer to some feature of the
  mutant phenotype
• Dickies small eye (Thieler et al., 1978, Anat
  Embryol (Berl), 155 81-86) is now Pax6
• Gleeful "This gene encodes a C2H2 zinc finger
  transcription factor with high sequence
  similarity to vertebrate Gli proteins, so we have
  named the gene gleeful (Gfl)." (Furlong et al.,
  2001, Science 293 1632)

Whats in a name!
Rose is a rose is a rose is a rose!
Gene Nomenclature

• Disease names
• Mobius Syndrome with Polands Anomaly
• Werners syndrome
• Downs syndrome
• Angelmans syndrome
• Creutzfeld-Jacob disease

• Accelerin
• Antiquitin
• Bang Senseless
• Bride of Sevenless
• Christmas Factor
• Cockeye
• Crack
• Draculin
• Dickies small eye

• Draculin
• Fidgetin
• Gleeful
• Knobhead
• Lunatic Fringe
• Mortalin
• Orphanin
• Profilactin
• Sonic Hedgehog

24
Rose is a rose is a rose is a rose.. Not Really!
What is a cell?

• any small compartment
• (biology) the basic structural and functional
  unit of all organisms they may exist as
  independent units of life (as in monads) or may
  form colonies or tissues as in higher plants and
  animals
• a device that delivers an electric current as a
  result of chemical reaction
• a small unit serving as part of or as the nucleus
  of a larger political movement
• cellular telephone a hand-held mobile
  radiotelephone for use in an area divided into
  small sections, each with its own short-range
  transmitter/receiver
• small room in which a monk or nun lives
• a room where a prisoner is kept

Image Sources Somewhere from the internet
25
Foundation Model Explorer
26

• COLORECTAL CANCER 3-BP DEL, SER45DEL
• COLORECTAL CANCER SER33TYR
• PILOMATRICOMA, SOMATIC SER33TYR
• HEPATOBLASTOMA, SOMATIC THR41ALA
• DESMOID TUMOR, SOMATIC THR41ALA
• PILOMATRICOMA, SOMATIC ASP32GLY
• OVARIAN CARCINOMA, ENDOMETRIOID TYPE, SOMATIC
  SER37CYS
• HEPATOCELLULAR CARCINOMA SOMATIC SER45PHE
• HEPATOCELLULAR CARCINOMA SOMATIC SER45PRO
• MEDULLOBLASTOMA, SOMATIC SER33PHE

The REAL Problems
Many disease states are complex, because of many
genes (alleles ethnicity, gene families, etc.),
environmental effects (life style, exposure,
etc.) and the interactions.
27
The REAL Problems
28
Methods for Integration

• Link driven federations
• Explicit links between databanks.
• Warehousing
• Data is downloaded, filtered, integrated and
  stored in a warehouse. Answers to queries are
  taken from the warehouse.
• Others.. Semantic Web, etc

29
Link-driven Federations

• Creates explicit links between databanks
• query get interesting results and use web links
  to reach related data in other databanks
• Examples NCBI-Entrez, SRS

30
http//www.ncbi.nlm.nih.gov/Database/datamodel/
31
http//www.ncbi.nlm.nih.gov/Database/datamodel/
32
http//www.ncbi.nlm.nih.gov/Database/datamodel/
33
http//www.ncbi.nlm.nih.gov/Database/datamodel/
34
http//www.ncbi.nlm.nih.gov/Database/datamodel/
35
Link-driven Federations

• Advantages
• complex queries
• Fast
• Disadvantages
• require good knowledge
• syntax based
• terminology problem not solved

36
Data Warehousing
Data is downloaded, filtered, integrated and
stored in a warehouse. Answers to queries are
taken from the warehouse.

• Advantages
• Good for very-specific, task-based queries and
  studies.
• Since it is custom-built and usually
  expert-curated, relatively less error-prone.

• Disadvantages
• Can become quickly outdated needs constant
  updates.
• Limited functionality For e.g., one
  disease-based or one system-based.

37
Algorithms in Bioinformatics

• Finding similarities among strings
• Detecting certain patterns within strings
• Finding similarities among parts of spatial
  structures (e.g. motifs)
• Constructing trees
• Phylogenetic or taxonomic trees evolution of an
  organism
• Ontologies structured/hierarchical
  representation of knowledge
• Classifying new data according to previously
  clustered sets of annotated data

38
Algorithms in Bioinformatics

• Reasoning about microarray data and the
  corresponding behavior of pathways
• Predictions of deleterious effects of changes in
  DNA sequences
• Computational linguistics NLP/Text-mining.
  Published literature or patient records
• Graph Theory Gene regulatory networks,
  functional networks, etc.
• Visualization and GUIs (networks, application
  front ends, etc.)

39
Disease Gene Identification and Prioritization
Hypothesis Majority of genes that impact or
cause disease share membership in any of several
functional relationships OR Functionally similar
or related genes cause similar phenotype.

• Functional Similarity Common/shared
• Gene Ontology term
• Pathway
• Phenotype
• Chromosomal location
• Expression
• Cis regulatory elements (Transcription factor
  binding sites)
• miRNA regulators
• Interactions
• Other features..

40
Background, Problems Issues

• Most of the common diseases are multi-factorial
  and modified by genetically and mechanistically
  complex polygenic interactions and environmental
  factors.
• High-throughput genome-wide studies like linkage
  analysis and gene expression profiling, tend to
  be most useful for classification and
  characterization but do not provide sufficient
  information to identify or prioritize specific
  disease causal genes.

41
Background, Problems Issues

• Since multiple genes are associated with same or
  similar disease phenotypes, it is reasonable to
  expect the underlying genes to be functionally
  related.
• Such functional relatedness (common pathway,
  interaction, biological process, etc.) can be
  exploited to aid in the finding of novel disease
  genes. For e.g., genetically heterogeneous
  hereditary diseases such as Hermansky-Pudlak
  syndrome and Fanconi anaemia have been shown to
  be caused by mutations in different interacting
  proteins.

42
PPI - Predicting Disease Genes

• Direct proteinprotein interactions (PPI) are one
  of the strongest manifestations of a functional
  relation between genes.
• Hypothesis Interacting proteins lead to same or
  similar disease phenotypes when mutated.
• Several genetically heterogeneous hereditary
  diseases are shown to be caused by mutations in
  different interacting proteins. For e.g.
  Hermansky-Pudlak syndrome and Fanconi anaemia.
  Hence, proteinprotein interactions might in
  principle be used to identify potentially
  interesting disease gene candidates.

43

• Prioritize candidate genes in the interacting
  partners of the disease-related genes
• Training sets disease related genes
• Test sets interacting partners of the training
  genes

44

• Example Breast cancer

15
342
2469
45
ToppGene General Schema
46
(No Transcript)
47
(No Transcript)
48
PubMed
OMIM
49
http//sbw.kgi.edu/