This document discusses the need for annotation of genomic data given the deluge of information from next generation sequencing. It outlines that clinical-grade annotation is important for application. Many sources of annotation are discussed, including databases, literature, testing labs, and crowdsourcing. However, it emphasizes that specialized human curation remains essential for high quality annotation.
Take control of your SAP testing with UiPath Test Suite
Genomic Data Annotation: Making Sense of the Deluge
1. One million monkeys with typewriters
Annotations of the Genomic Data Deluge
Genome Informatics Alliance
Portland, 28/29 March 2012
Dr. Frank Schacherer, CTO, BIOBASE GmbH
frank.schacherer@biobase-international.com
2. Disclaimer: no actual monkeys
involved
In 2003 the Arts Council for England
paid £2,000 for a real-life test of the
theorem involving six Sulawesi crested
macaques, but the trial was abandoned
after a month.
AT C
G
G AT TT The monkeys produced five pages of
TT A text, mainly composed of the letter S,
C
GTA CG but failed to type anything close to a
CGC word of English, broke the computer
G
G TA C and used the keyboard as a lavatory.
A
ATA
C
TTG A A http://www.telegraph.co.uk/technology/news/8789
C
TG G 894/Monkeys-at-typewriters-close-to-reproducing-
C
CGT AT Shakespeare.html
T
4. A deluge of data
• deluge (plural deluges)
– A great flood or rain.
The deluge continued for hours,
drenching the land and slowing traffic
to a halt.
– An overwhelming amount of
something.
The rock concert was a deluge of
sound.
5. Media perception
Science 2011
The Power Of Digitizing
Health Affairs 2009
Human Beings
17 Feb 2012
Soon, $1,000 Will Cost of Gene Sequencing
Map Your Genes Falls, Raising Hopes for
10 Jan 2012 Medical Advances
'Personalized Medicine'
7 March 2012
Hits a Bump / March 2012
6. Life cycle of data annotation
Understan Derive
dMap Analyze
Annotate Publish
Rank Curate
7. How to predict mutation effects
• Overlap with other data
– dbSNP, 1000 genomes
– Relatives and Controls
• Algorithmically
– Frameshift, Nonsense, Stop
gain/loss, Non-synonymous
changes (SIFT, PolyPhen, ...)
• Based on annotation
– known functional regions
(active sites, binding sites, ...)
• Directly known effects
– HGMD
Bioinformatics, Vol. 26 no. 16 2010, pages 2069; 10.1093/bioinformatics/btq330
9. What data do we need for clinical
application
ACCE takes its name from the four main criteria for evaluating a genetic test —
analytic validity, clinical validity, clinical utility and associated
ethical, legal and social implications
Centers for Disease Control and PreventionOffice of Public Health Genomics (OPHG)
10. Ideal Annotation for clinical use?
• Variants N=12
– Pathogenic, Uncertain, Benign 4 Testing
(Clinical Validity,Who/When, Methods,
– Severities, if known
Interpretation, Cost)
– Ethnicities/Frequencies 4 Management,
– Number of cases Clinical Significance, Implications
– Symptoms In conjunction with 3 Actionability, Clinical Utility
other mutations 3 Clinical manifestations
• ( Pathophysiology, Phenotype, Prognosis,
Evidences
Severity, Penetrance,
– Not weighted equally Pleiotropy)
– Risks of incorrect classification 2 Frequency
not equal between genes (especially indicate most common variants)
2 Inheritance and
Data from: Howard P. Levy, MD, PhD
Johns Hopkins University
de novo mutation rate
2 Evidence-based
Data from: Elaine Lyon, Ph.D.,
FACMG University of Utah & 1 Clinical Decision Support in EHR
ARUP Laboratories
11. Who provides annotation?
Payor Test Lab Curator Researcher
Patient MD/Geneticist Anybody Computer
12. Surveys & Patient Self-annotation
nature biotechnology VOLUME 29 NUMBER 5 MAY 2011
Knaus, William A.
BUILDING A GENOME Patients with serious diseases may experiment with drugs that have
ENABLED ELECTRONIC not received regulatory approval. Online patient communities
MEDICAL RECORD structured around quantitative outcome data have the potential to
provide an observational environment to monitor such drug usage
and its consequences. Here we describe an analysis of data
reported on the website PatientsLikeMe by patients with amyotrophic
lateral sclerosis (ALS) who experimented with lithium carbonate
15. Testing Lab data
A safe and secure route for sharing variant data
The Diagnostic Mutation Database (DMuDB) is a unique repository of high
quality variant data collected from accredited clinical genetic testing
laboratories in the UK National Health Service (NHS).
It provides a safe and secure way for variant data to be shared within and
between laboratories in order to support safer, more consistent
diagnoses. The database was established in order to address the lack of
data-sharing or publication in the genetic testing community.
DMuDB is used regularly by genetic scientists:
• to check a new variant against existing reported variants from
other laboratories
• to check for co-reported variants
• as a part of regular re-assessment of unclassified variants
• via the Universal Browser as part of complex searches
covering multiple databases
www.ngrl.org.uk/Manchester
18. Crowdsourcing reality
…biological databases can be
“The future of curated by a diffuse network of
biocuration
To thrive, the field that
volunteers? This is certainly not the
links biologists and their case and at the core of every
data urgently needs successful wiki database are a group
structure, recognition
and support. “
of dedicated experts who do the bulk
NATURE|Vol 455|2008 of the data curation.
20. Data Annotation Professionals
• Clear incentives
• Background in life sciences (MSc/PhD)
• Curation is sole focus of work
• Knowledge of standards, databases, formats,
specialized tools
Huge volumes of primary data are currently
archived in numerous open-access databases, and
with new generation technologies becoming more
common in laboratories, large datasets will become
even more prevalent than today. The lasting
archiving, accurate curation, efficient analysis and
precise interpretation of all of these data are a
challenge. Collectively, database development and
biocuration are at the forefront of the endeavor to
make sense of this mounting deluge of data.
25. Conclusions on annotation
• Clinical-grade annotation may be the most
important task ahead
• NGS itself contributes to generate evidence
• Many different sources and ways of annotation
exist
• Human, specialist annotation remains essential
(monkeys nonwithstanding)
26. • BIOBASE Employees all around the world
• David Cooper, University of Cardiff
Thank you!
• Andrew Deveraux, NGRL
• Patrick Willems, MutaBase
• Johan den Dunnen, HVP & Leiden University Medical Center
• Anthony J. Brooks, GEN2PHEN & University of Leicester
• Samir K. Brahmachari , OSDD
Gene Regulation Analysis Human Mutation & Functional Analysis
Variant Analysis
sales@biobase-international.com
www.biobase-international.com
Hinweis der Redaktion
Callback to last year in Verona, where shakespeares romeo and julia played „ An infinite number of monkes with typewriters (or one monkey with infinite time) in principle would be able to write all the works of shakespeare“ the idea of getting things right by throwing impractically large resources at them (the monkeys would take much longer than the length of the universe) How can we deal with millions of genomes, how can we annotate them, facing the same limitation in resources Goes back to Aristoteles in regard to permutations, and more recently 1913 — Émile Borel’s essay — “Mécanique Statistique et Irréversibilité”
Datas not bad, ist only bad if we do not know what to do with it
„ Where is the wisdom we have lost in knowledge? Where is the knowledge we have lost in information?“ – T.S. Eliot
Analytic validity How accurately and reliably the test measures the genotype of interest. Clinical validity How accurately the test detects or predicts the outcomes of interest. Clinical utility How likely the test is to significantly improve patient outcomes.
Facebook for genomes Ebay for genomes?
Bad fít for whole genome/exome Quality consistency issues
Wikipedia = Crowdsourcing‘s Posterchild for distributed curation: Everybody can contribute quality killed Britannica Linus‘ Law: "given enough eyeballs, all bugs are shallow“ -- Eric S. Raymond, named in honor of L. Torvalds India: Incentive of getting published
Has not worked out that well in practice, for biology/science because of researchers spending time rather to do research and get published, and because of difficulites with maintaining standards Ideas: Force journals mandate submission of data into databases journals require gene symbols, accessions, etc Tie career advancement to annotation with Microattribution Crowdfunding Suggested solutions: Force ‘em? journals mandate submission of data into databases authors provide a machine-readable XML summary journals require gene symbols, accessions for genes and isoforms, description of species, cell types, genotypes Tie career advancement to annotation? Author IDs, Microattribution Fund community annotation (crowdfunding)? Valuable suggestions to enable many tools. For lowest level of data curation. Changing career eval would mean changing the entire credit system for research which sits on peer reviewed author papers
Use expert who are paid to collect information manually into databases
We train curators for up to half a year before they do live curation