Course Objectives

1
Course Objectives

• To learn about research studies driving the field
  and computing techniques that have been
  developed.
• To learn about computational and informatics
  projects related to biology, medicine, and other
  life science disciplines at Emory.
• To learn about opportunities for summer research
  and dissertation topics.
• To stimulate ideas for further collaboration
  between Math/CS and X.

But impossible to give a complete treatment of
field.
2
Why?Computational and Life Science?

• So you need to go see a doctor?

3
Why CLS?

• A look at your personal medical history
• Do you eat right? Do you exercise?
• Do you smoke? Do you drink alcohol?
• What is your current / past profession?
• Have you had any of the following
• Difficulty breathing Circulatory problems
• Eating disorder Behavioral issues
• Heart problems Visual impairment
• Allergies Asthma

4

• A shallow picture of your medical profile.

Ancestral Knowledge Future symptoms Information
today Disease
5

• Follow a specific problem (nausea)
• Additional lab tests (bacterial, viral, hormonal,
  ulcers, celiacs, diabetes, cancer)
• More specific questions to determine extent of
  problem and other symptoms

6

• Look at your family medical history
• Has any one in your immediate family had any of
  the following
• Heart disease Diabetes
• Cancer Alzheimers
• If so who?
• Mother, Father, Siblings, Aunts/ Uncles,
  Grandparents

7
Ancestral Knowledge Future symptoms Information
today Disease
8

• But how can this picture give any facts of the
  specifics, or causes of diseases within your
  ancestral medical history.
• Was your grandmother a heavy smoker, was your
  grandfather overweight.
• Even if similar symptoms the causes may be more
  due to personal choices or environment
• How could we decipher the facts / causes?
• (venn diagram of symptoms and causes)

9

• A deeper picture of your medical profile.
• More depth.
• Cumulation of information points to specific
  diagnosis.
• But this required symptoms

10

• Current diagnostics like to follow a single path
  at a time.
• Do test, examine results, prove or disprove.
• If disproved, evaluate a second route.
• Efficient in the case of clinical costs,
  inefficient in the cost of time.

11

• Even with symptoms the diagnosis may be wrong.
• Car link

12

• Well, the flow chart diagnosis has been completed
  and the final result is a defective PCM. I just
  had a strange feeling and I just cannot seem to
  accept that.
• The other PCM made no difference.
• What went wrong? The diagnostic trouble chart was
  carefully followed and yet the end result was
  incorrect? Was the flow chart misleading?
  Absolutely NOT, one thing to KEEP IN MIND when
  following the flow charts is that the "MOST
  LIKELY" cause will be shown. There is no way to
  know exactly what fails from one case to another.
  I don't fault the information at all, as a matter
  of fact, even though the problem was not yet
  known, I do know, by following the flow chart,
  what areas are correct.
• So, there you have it, not every cause will be
  listed as the end result when using a diagnostic
  flow chart.

13

• The current information is not sufficient
• What if we could add to this information?
• What would you want to add?
• Can we start the diagnosis earlier
• before symptoms?
• Is ones personal prevalence for a specific
  disease measurable?
• How would one determine this?

14

• What can be measured?
• Recorded?
• Compared between normal and diseased?
• Can a variance be measured?
• Is this variance predictive?

15

• Back to data points.
• Clinical lab studies (images, chemical
  monitoring, physical exam, etc.)
• Scientists are currently accumulating data in
  multiple areas (DNA, RNA, protein, etc.)
• Recording data for normals, diseased, with
  treatment, without treatment.
• Many, many replicates!
• Billions of data points
• Comparison
• What features correlate with normal or disease,
  etc.
• Can this feature be predictive?

16
Technology and CS Requirements

• Given 1000s of instances
• queriable database
• feature definition, feature extraction
• feature selection
• comparison, classification, correlation
• prediction
• modeling predictive risk models

Will discuss this protocol in many different
instances.
17
DILS 2005 keynotes

• Shankar Subramaniam, Professor of Bioengineering
  and Chemistry at UCSD
• the standard paradigm in biology hypothesis to
  experimentation (low throughput data) to models
  is being replaced by data to hypothesis to
  models and experimentation to more data and
  models.
• need for robust data repositories that allow
  interoperable navigation, query and analysis
  across diverse data, a plug-and-play environment
  that will facilitate seamless interplay of tools
  and data and versatile biologist-friendly user
  interfaces.

18
Databases

• Data, Data, Data
• Organization of database (studies, experiments,
  sample sets, patients, treatments)
• Meta-data, including experimental conditions and
  clinical data
• repeated data points
• Secondary experimental procedures (more variate
  data)
• Incomplete data sets
• Multiple analysis runs (multiple data sets)
  (scaling, normalization, archive, comparisons,
  requerying)
• From experimental results, re-query data on other
  meta-data and reprocess
• Annotations of experimental data points (genes,
  proteins, etc.)

19
Technology and CS Requirements

• Definition of data structure
• Download of data into database
• Storage and retrieval
• Security
• Integrated database, data archive, analytical
  results archive
• ...
• Feature selection and modeling
• generation of sophisticated, integrated
  predictive risk models

20
Predictive Health

• Health general condition of the body or mind
  with reference to soundness and vigor, freedom
  from disease or ailment.
• Diagnose to recognize (a disease) by signs and
  symptoms, to analyze the cause or nature of.
• Predict to declare in advance (of symptoms) on
  the basis of observation, experience, scientific
  reasoning.

21
Predictive Health

• Predictive health is an emerging paradigm that
  emphasizes maintaining health by detecting the
  genetic risk factors for illness and taking steps
  to prevent disease or illness before it starts.
  
• In the future, providers will combine an
  individuals genetic information with cutting
  edge biotechnology to keep that person healthy.
  Eventually, the occurrence of disease will be
  seen as a failure of the health care system,
  rather than its main focus.
• Momentum Summer 2006, Seeking Ponces Dream

22
Momentum Winter 2006-2007, DNA Rubric

• SNP accounts for some of the variation among
  humans. These naturally occurring differences,
  polymorphisms, help explain difference in human
  appearance and why some people are susceptible to
  diseases like lung cancer and others arent.
  They also provide an explanation for why there
  can be individualized responses to environmental
  factors and medications.
• These patterns (of specific variation) will help
  us predict the future health of an individual and
  develop personalized health treatments, including
  specific drugs tailored to each individual, given
  their specific genetic code.
• Scott Devine, PhD, Biochemistry

23
Predictive Health 2007

• Center for Health Discovery and Well-Being
• participants - 100 - 200 generally healthy people
• collect physical, medical and lifestyle
  histories, environmental factors
• perform 50 blood and plasma tests (including
  genotypes) that target known critical predictors
  of health and illness
• the research program will develop and validate
  novel biologic markers to predict health, disease
  risk, and prognosis.
• based on these profiles and a predictive risk
  model, each participant will be prescribed a
  personalized health program designed to address
  individual risks.

24
Technology and CS Requirements

• Database and Security
• Integrated database and data archive
• Feature definition, feature extraction
• Feature selection
• Comparison, classification
• Prediction
• Modeling sophisticated, integrated predictive
  risk models
• Annotations, data-mining
• ...

25
Systems Biology

• Systems Biology is the science of discovering,
  modeling, understanding and ultimately
  engineering at the molecular level the dynamic
  relationships between the biological molecules
  that define living organisms.
• Leroy Hood, ISB
• http//www.systemsbiology.org/Systems_Biology_in_D
  epth

26
Momentum Winter 2006-2007, Fresh Air

• Molecular signaling pathways within normal cells
  follow a cascade of molecular reactions that emit
  proteins, which turn on
• The premise acknowledges that a single genetic
  mutation doesnt cause lung cancer. Instead
  there are many causes on the cellular level, with
  many genetic mutations from many different
  sources.
• Fadlo Khuri, PhD.
• Clinical and Translational Research

27
(No Transcript)
28

• Ingenuity

29

• List of Model Repositories
• CellML biochemical and cellular processes
• DOQCS DB of Quantitative and Cellular
  Signalling
• Model DB Sense Lab, nerves and neurons
• SigPath and SigMoid Signalling pathways
• PathArt Metabolic pathways

30

• Systems biology markup language

31
Technology and CS Requirements

• Database and Security
• Integrated database and data archive
• Feature definition, feature extraction
• Feature selection
• Comparison, classification, correlation
• Prediction
• Modeling sophisticated, integrated predictive
  risk models
• Annotations, data-mining
• ...

32
CS in CLS?

• 5 of biological researchers have hired a CS
  or DB staff.
• 95 who dont because
• do not see the need,
• have no experience in CS or managing CS,
• can not raise the funds.
• Communication, Communication, Communication

33
Meta-Objectives

• How does a CS knowledgeable person become an
  X-informatics or computational-X researcher?
• How useful is it to work with just symbolic
  abstractions?
• How much X does one need to learn for the
  research to be meaningful?
• How can it be more mutual collaboration?
• Most of the time, it is just CS servicing X.
• X researchers really dont care how the CS is
  done. Just Do It!

34
Meta-Objectives CS in CLS?

• The CS scientist should know enough biology to
  probe beyond the obvious question that the
  biologist is asking.
• Be able to and willing to offer direction. You
  can use this CS technology or algorithm to answer
  X about your data.

35
NCBI Derivative Sequence Data (Maureen J. Donlin,
St. Louis University)
C
C
Curators
GA
GA
ATT
C
GA
GA
ATT
C
RefSeq
TATAGCCG
ACGTGC
TATAGCCG AGCTCCGATA CCGATGACAA
ATTGACTA
CGTGA
TTGACA
Labs
TTGACA
TTGACA
ACGTGC
Genome Assembly
TATAGCCG
ACGTGC
TATAGCCG
ATTGACTA
CGTGA
CGTGA
ATTGACTA
TATAGCCG
CGTGA
ATTGACTA
TTGACA
ATTGACTA
TATAGCCG
ATTGACTA
TATAGCCG
TATAGCCG
TATAGCCG
TATAGCCG
ATT
C
GenBank
GA
UniGene
AT
GA
C
C
Algorithms
ATT
C
C
GA
GA
ATT
GA
GA
ATT
C
C
GA
ATT
GA
GA
ATT