Bioinformatics, Translational Bioinformatics, Personalized
Medicine Uma Chandran, MSIS, PhD Department of Biomedical
Informatics University of Pittsburgh [email protected] 412-648-9326
07/17/2013 Slide 2 Outline of lecture What is Bioinformatics?
Examples of bioinformatics Past to present What is translational
bioinformatics? Personalized Medicine Bioinformatics and
Personalized Medicine Slide 3 What is Bioinformatics?
http://en.wikipedia.org/w iki/Bioinformatics
http://en.wikipedia.org/w iki/Bioinformatics Application of
information technology to molecular biology Databases Algorithms
Statistical techniques Slide 4 Bioinformatics examples Sequence
analysis Genome annotation Evolutionary biology Literature analysis
Analysis of Gene Expression Analysis of regulation Analysis of
protein expression Analysis of mutations in cancer Comparative
genomics Systems Biology Image analysis Protein structure
prediction From Wikipedia Slide 5 Early Bioinformatics Robert
Ledley and Margaret Dayhoff First bioinformaticians Using IBM 7090
and punch card analyzed amino acid structure of proteins Created
amino acid scoring matrix Protein evolution Protein sequence
alignment http://blog.openhelix.eu/?p=1078 Slide 6 Sequence
analysis Databases to store sequence info Phage -X174 sequenced in
1977 GenBank 30, 000 organisms 143 billion base pairs BLAST program
for sequence searching Algorithms, databases, software tools Slide
7 Evolutionary biology Compare relationships between organism by
comparing DNA sequences Now whole genomes Can even find single base
changes, duplication, insertions, deletions Uses advanced
algorithms, programs and computational resources Slide 8 Literature
mining Millions of articles in the literature How to find
meaningful information Natural language processing techniques
Example Type in p53 or PTEN in Pubmed will retrieve 1000s of
publications How to summarize all the information for a particular
gene Function, disease, mutations, drugs IHOP database creates
network between genes and proteins for 30000 genes Slide 9 Genome
annotation Marking genes and other features in DNA Algorithms,
software Slide 10 Bioinformatics Interdisciplinary discipline
Gene/proteins/function/ - Biologist In Cancer
Physician/Scientist/Biologist Algorithms, for example, BLAST
Math/CS Separate Signal from Noise, Diff gene expression,
correlation with disease Statistician Tools, Software, Databases
Software developers, programmers Aim to make sense of biological
data Slide 11 Translational bioinformatics Translational =
benchside to bedside Bringing discoveries made at the benchside to
clinical use the development of storage, analytic, and interpretive
methods to optimize the transformation of increasingly voluminous
biomedical data into proactive, predictive, preventative, and
participatory health. Translational bioinformatics includes
research on the development of novel techniques for the integration
of biological and clinical data and the evolution of clinical
informatics methodology to encompass biological observations. The
end product of translational bioinformatics is newly found
knowledge from these integrative efforts that can be disseminated
to a variety of stakeholders, including biomedical scientists,
clinicians, and patients. Translational = benchside to bedside Atul
Butte, JAMIA 2008;15:709-714 doi:10.1197 Slide 12 Central dogma DNA
is transcribed to RNA RNA is translated to protein Many regulatory
processes control these steps Slide 13 Molecular Biology Primer 20,
000 genes Many transcripts, many proteins More than 20, 000
proteins Southern, Northern, Western Blots Slide 14 Biological
questions DNA Are there any mutations sickle cell anemia Cystic
fibrosis Hemophilia Other diseases such as diabetes, cancer ??
Polymorphisms Variation in the population Mutation Slide 15 DNA
amplification Are there regions of amplification or deletions that
correlate with disease If so, what genes are present in these
regions HER2 amplification in breast cancer EGFR mutations in lung
cancer Slide 16 RNA DNA is transcribed to RNA Approximately 20K
genes RNA levels will differ in different conditions Liver, kidney,
cancer, normal, treatment etc Diagnosis or prognostic microRNAs
level lnncRNAs Splicing differences mRNA Slide 17 Clinical
questions DNA level Are there mutations or polymorphism between
different cancer patient groups Good outcome v bad outcome Early
stage vs late stage Therapy responders v non-responders Examples:
Renal cell, prostate cancer etc RNA Are there specific transcripts
mRNA, microRNA - that are up or down and are signature for outcome,
disease and response 1000s of studies Consortia projects TCGA The
Cancer Genome Atlas projects Profile 500 samples of each cancer for
DNA, RNA changes Slide 18 Slide 19 Slide 20 Molecular Biology
Primer 20, 000 genes Many transcripts, many proteins More than 20,
000 proteins Southern, Northern, Western Blots Slide 21 Base
pairing Microarray and Northern/Southern blots Exploit the ability
of nucleotides to hybridize to each other Base pairing
Complementary bases A :T (U) G: C Slide 22 Northern Sensitivity and
dynamic range low Slide 23 How are these changes measured Example:
Northern blot (measure RNA)
http://www.youtube.com/watch?v=KfHZFyADnNg
http://www.youtube.com/watch?v=KfHZFyADnNg Workflow of Northern
blot Key points mRNA run on gel separated by size transferred to a
membrane immobilized Have a hypothesis for example studying RNA
level for BRCA in normal and cancer Only probe for a mRNA or
transcript is labeled or tagged probe is prepared and labeled with
radioactivity Hybridized to X-ray film Only that mRNA is detected
and quantitated Slide 24 Microarrays Solid surface Many different
technologies Affy, Illumina, Agilent Probes are synthesized on the
solid surface Synthesized using proprietary technology Probe are
selected using proprietary algorithms RNA (or DNA) is in solutions
RNA is labeled or tagged Hybridized to the chip Tagged RNA is
quantitated Compare between conditions Slide 25 Affymetrix Slide 26
Need for computational methods Data Management Each file for a chip
experiment is large 100MG x 10 = 1G Generates Gigabytes of data
Data preprocessing Convert raw image into signal values Data
analysis 1000s of genes (or SNPs) and few samples How to find
differences between samples What statistical methods to use? Like
finding needle in a haystack Slide 27 How to analyze? Samples
GENESGENES Normal Tumor Noise reduction Background subtraction
Normalization Data Analysis Slide 28 Data analysis Class discovery
Are there novel subclasses within data? Class comparison How are
tumor and normal different in expression? Which SNPs are different?
Class prediction Predict class of new sample Advanced pathway
Analysis Slide 29 Pathway Analysis Slide 30 Analytic methods many
studies, many methods Dupuy and Simon, JNCI; 2007 Slide 31 SNPs to
detect Copy Number changes diploid deletion amplification Slide 32
Hagenkord et al; Modern Pathology, 21:599 Slide 33 What is
personalized medicine Personalized medicine is the tailoring of
medical treatment to the individual characteristics of each
patient. Based on scientific breakthroughs in understanding of how
a persons unique molecular and genetic profile makes them
susceptible to certain diseases. ability to predict which medical
treatments will be safe and effective for each patient, and which
ones will not be. From ageofpersonalizedmedicine.org Slide 34
Personalized Medicine From ageofpersonalizedmedicine.org Slide 35
Personalized Medicine From Fernald et al; Bioinformatics, 13: 1741
Slide 36 Examples of personalized medicine Breast cancer 30% of
patients over express HER2 Treated with Herceptin Oncotype Dx: gene
expression predicting recurrence Cardiovascular Patients response
to Warfarin, the blood thinner Response determined by polymorphism
in a CYP genes Slide 37 Personalized Medicine Examples of
personalized medicine resulted from studies that generate Lots of
data Rely on bioinformatics methods to discover these associations
Oncotype Dx: Gene expression studies of large number of patients
CYP polymorphisms Discover single nucleotide polymorphisms in
patient polulations and association with response Initial studies
done with PCR methods Slide 38 Personalized Medicine Current
examples are few in numbers Making personalized medicine a reality
Generate the data Discover the associations Find targeted therapies
Genome sequences prices are dropping Large scale genome information
is coming: 1000 genome TCGA ICGC Also possible to commercially
sequence a persons genome Processing all this data into translating
these discoveries into medical practice has many challenges Slide
39 Bioinformatics challenges in personalized medicine Processing
large scale robust genomic data Interpreting the functional impact
of variants Integrating data to relate complex interactions with
phenotypes Translating into medical practice Fernald et al;
Bioinformatics: 13: 1741 Slide 40 Era of Personalized medicine
Shift from microarrays to Next Gen Sequencing Slide 41 Central
dogma DNA is transcribed to RNA RNA is translated to protein Many
regulatory processes control these steps Slide 42 Next Gen
Sequencing Directly sequence DNA to determine SNP CN Expression,
mRNA, microRNA Protein binding sites Methylation Initial steps
depend not on hybridization but also on base pairing or
complementarity and DNA synthesis Bioinformatics is extremely
challenging Slide 43 Next Gen Sequencing Slide 44 NGS in
personalized medicine Whole genome sequencing Sequence genomes and
find variants (1000 genome project) Find variants associated with
disease phenotype Sequence exomes only Find coding region variants
associated with phenotypes RNA seq RNA sequence signatures
associated with phenotype Slide 45 Microarrays v NGS RNA Seq
Restricted to probes on chips Only transcripts with probes File
sizes in MBs to GB Algorithms, methods Typically done on PCs
Storage on hard drives No predetermined probes Can detect
everything that is sequenced More applications than microarray Very
large file sizes Computationally very intensive Clusters,
supercomputers Large scale storage solutions Slide 46 Microarrays v
RNA seq Expression Analysis Dynamic range is low Statistic to
determine expression based on signal Many methods in the last 10
years Dynamic range is high Based on reads Statistics based on
counts Affected by read length, total number of transcripts, lack
of replicates Slide 47 Read mapping Alignment Denovo assembly
Mapping to reference genome Based on complementarity of a given 35
nucleotide to the entire genome Computationally intensive Million
of 35 bp reads has to search for alignment against the reference
and align spefically to a given regions Large file sizes Sequence
files in the TB Aligned file BAM files Several hundred GB Reference
genome Slide 48 Sequence variation Slide 49 Bioinformatics
challenges in personalized medicine Processing large scale robust
genomic data Suppose we want to identify DNA variants associated
with disease Which technology How much data How to analyze the data
How to identify variants Each genome can have millions of variants
300, 000 new variants i.e, not in existing databases Will have to
separate error from true variants 1 error per 100 kb can lead to
30,000 errors in a single experiment Why do these errors happen?
Fernald et al; Bioinformatics: 13: 1741 Slide 50 Bioinformatics
Challenges Data Which technology to use Each technology has
different error rates, Ion Torrent (higher error rate), SOLID,
Illumina Speed of generation of data Ion Torrent is faster
Application Whole genome or exome or targeted exome Analysis
Algorithms, speed, accuracy BLAST is not good for WGS Other new
algorithms Speed of analysis Alignment can take days Alignment
relies on matches between sequence and reference genome How much
mismatches to tolerate True mismatch or error sequencing error,
true mismatch is it a SNP Quality of reference genome Large amounts
of data Each whole genome sequencing experiment can generate TB of
data Where to store patient privacy Servers, locations, networking
Sample sizes how many samples to sequence to discover the
association with disease Slide 51 Bioinformatics Challenges
Technology Ion Torrent, SoLiD, Illumina Each has its own error
rates Speed of data generation Dependent on application WGS or
exome Analysis Algorithms, speed, accuracy Speed of analysis
Alignment can take days Alignment relies on matches between
sequence and reference genome How much mismatches to tolerate True
mismatch or error sequencing error, true mismatch is it a SNP
Quality of reference genome Slide 52 From Mark Boguskis
presentation at the IOM, July 19, 2011 Slide 53 Slide 54 Slide 55
Molecular Diagnostics using NGS From Mark Boguskis presentation at
the IOM, July 19, 2011 Slide 56 NGS Bioinformatics - medicine
Infrastructure Storage, backup, archive Where HIPAA compliant?
Network How to move data Analysis Methods statistics, annotation
Computing resources How many samples can be handled at a time? Time
to report Slide 57 NGS and bioinformatics Slide 58 Next Gen
Sequencing From Mark Boguskis presentation at the IOM, July 19,
2011
