
HST 952	Computing for Biomedical Scientists	Fall 2003

Course Information


Course announcements Course calendar Level and Units: Grad H; 3 (lecture)-0 (lab)-9 (prep) Time and Place: Tuesdays/Thursdays 11 a.m. - 12:30 p.m. Location: MEC Rm 334 Tosteson Medical Education Center Harvard Medical School (260 Longwood Avenue) Exception: MEC, Room 333 on Sept. 4 and Sept. 25 Instructors: Omolola Ogunyemi, Aziz Boxwala, Qing Zeng For general correspondence and homework submission send e-mail to ta952@dsg.harvard.edu cc: rclacson@mit.edu Office Hours: Tuesdays 10-11 a.m. as requested Course mailing list: hst952@yahoogroups.com Textbooks (required): Java: An Introduction to Computer Science and Programming, 3rd edition Author: Walter Savitch Prentice Hall, 2003 Modern Database Management, 6th edition Authors: Fred R. McFadden, Jeffrey Hoffer, Mary B. Prescott Addison-Wesley, 2001 Supplementary texts/readings: Foundations of Computer Science Authors: Alfred Aho, Jeffrey Ullman W. H. Freeman and co., 1995 Handbook of Medical Informatics Authors: Jan H. van Bemmel, Mark A. Musen Springer-Verlag, 1997 Principles of Database and Knowledge-Base Systems, Volume 1 Author: Jeffrey D. Ullman W H Freeman & Co., 1988 Relevant papers or readings selected by instructors from: Knowledge Representation: Logical, Philosophical, and Computational Foundations Author: John F. Sowa Brooks/Cole Pub Co., 1999 Note: The textbooks/readings listed will be on reserve at Harvard's Countway Library of Medicine. Some of the texts can also be found at MIT's Barker Engineering Library, or at Harvard's Gordon McKay Applied Sciences Library.

Course description

This course introduces abstraction as an important mechanism for problem decomposition and solution formulation in the biomedical domain, and examines computer representation, storage, retrieval, and manipulation of biomedical data. As part of the course, we will briefly examine the effect of programming paradigm choice on problem-solving approaches, and introduce data structures and algorithms. We will also examine knowledge representation schemes for capturing biomedical domain complexity and principles of data modeling for efficient storage and retrieval. The final project involves building a medical information system that encompasses the different concepts taught in the course.

Computer science basics covered in the first part of the course are integral to understanding topics covered in the latter part, and for completing the assigned homework.

Goals:
With this course, we hope to provide a foundation for scientists interested in using computers for solving biomedical problems. Computing with biomedical data poses unique challenges with respect to data volume, complexity, and uncertainty in data and in domain knowledge. Students taking this course should come away with a grounding in abstraction for problem decomposition and solution formulation, data modeling, and information management. The latter are key to analysis, development, and proper design of information systems.

Prerequisites:
Biomedical background and an interest in computing

Programming will be done in Java; no prior familiarity with Java is assumed.

Topics

The following is an outline of topics to be covered in this course.

Part 1: Introduction to Computing

(11 classses)

Omolola Ogunyemi

Overview: How it all fits together
Solving problems with the computer
- Abstraction: creating a model for a problem and devising appropriate methods (algorithms) that can be automatically applied to solve the model problem
- Computational process: the solution to the model problem (and its subproblems)
- Data: the information processed by a computational process
- Computer program: a description of a computational process
- Programming language: the notation used for describing the computational process
Algorithmic programming paradigms
- Imperative programming
  - Programming is issuing commands (imperatives)
  - Variables and assignment are used to change state, and serve as an analog of the concept of a modifiable store supported by real computers
  - Lends itself easily to procedural and object-oriented programming styles
  - Examples of programming languages that naturally support this paradigm: C, C++, Java, Pascal, Fortran
- Declarative programming
  - Programming is defining/declaring a solution
  - Notion of state that reflects the underlying computer architecture is not an important consideration
  - Functional and logic programming are declarative programming approaches
  - Examples of programming languages that support this paradigm: Lisp, Scheme, ML, Prolog
Imperative Programming: Approaches for abstracting model state (and defining program structure)
- Procedural Programming
  - Commands are expressed as procedures/functions
  - Distinctions are maintained between data (state) and operations on data (behavior)
  - To carry out a task, issue a sequence of appropriate commands -- find/create appropriate procedures that modify data (state) and call them
- Object-Oriented Programming
  - Data (state) and functions that manipulate the data (behavior) are grouped together into a modular unit called an object
  - Commands are issued by objects
  - A task is carried out by creating a structured network of the appropriate objects and having the objects interact by exchanging messages
Object-Oriented Programming concepts
- Encapsulation
- Polymorphism
- Inheritance
Java language details
- Data types
- Expressions
- Control statements
Computational Processes and Abstraction
- Procedural Abstraction
  - Recursion
  - Iteration
- Data Abstraction
  - Lists
  - Binary trees
  - (Brief) Algorithms for searching
  - (Brief) Algorithms for sorting
  - (Brief) Complexity of algorithms (Big-Oh notation, etc.)

Part 2: Data and Knowledge Representation

Qing Zeng

Domain Ontology
1. Definition
  1. Ontology = catalog of types that exist in a domain D from perspective of some person using language L
  2. Informal ontology = types undefined or defined in natural language
  3. Formal ontology = types organized into partial order (lattice)
  4. Axiomatized formal ontology = axioms provide additional contraints on types
  5. Prototyped formal ontology = prototypical individuals provide characteristics. Similarity based on distance functions.
2. Motivation: need to represent medical data/knowledge
  1. Example of a patient case
  2. Intended applications
    - Patient care (Reporting, Summary, trend)
    - Financial, Manangement, Clinical Research, expert system, information retrieval, vocabulary discovery
  3. Difficulty in representation
    - Lack of consensus
    - Comprehensiveness
      - Drugs, lab tests, signs, symptoms, narrative reports
    - Cost
    - Proprietary interests
3. Distinction
  1. Definition
  2. Seven distinctions (Peirce and Whitehead)
    - Abstract/Physical
    - Independent/Relative/Mediating
    - Continuant/Occurrent
4. Trees
5. Axioms
6. Unified Medical Language System
7. Conceptual graphs
Logic
1. Propositional Logic
  1. Variable
  2. Connective
  3. Boolean algebra
  4. Tautology and deduction
2. Predicate Logic
  1. Statement
    - Predicate
    - Arguments
  2. Quantifiers
Special topics
1. Natural language semantics
  1. Language Analysis
    - Morphology
    - Syntax
    - Semantics
  2. Concepts and Relations
  3. Resolving Ambiguities
2. Uncertainty
  1. Source of uncertainty
  2. Representation Schemes
    - Fuzzy logic
    - Probability
    - Non monotonic logic…
  3. Example: Temporal Granularity as in TSQL

Part 3: Data management, querying, and retrieval

Aziz Boxwala

Data concepts and modeling
1. Nature of data
  1. Proxies and reality, metadata, data
  2. Relationship among data
2. Data models
  1. User models, conceptual models, physical models
  2. Modeling the meaning of data
    - Degree, dependencies, time, uniqueness, generalization-specialization, aggregation
  3. Modeling methods
    - ER models, Relational models, Object-oriented models, Hierarchical and network models
Relational model
1. Maintaining integrity of data
  1. Uniqueness and Keys
    - Primary and foreign keys
  2. Domain of data
  3. Normalization of models
    - Functional dependencies among data
    - Normal forms
2. Implementing a relational database
  1. Transforming logical models to physical models
    - Tables and views
    - Case tools
  2. Data dictionaries
  3. Joins and queries
    - Inner and outer joins
    - Query optimization
  4. SQL
    - DDL, DML, DQL
  5. Security models in RDBMS
Special topics
1. Overview of object-oriented data management
  1. Review of classes and objects, inheritance, encapsulation
  2. Object-relational DBMS
  3. Object-oriented databases
2. Modeling for analytical processing of data
  1. Star-join schema
    - Comparisons with transactional models

Assignments, Exams, and Grading

There will be weekly homeworks, consisting of programming assignments in Java. Assignments are generally due one week after they are distributed. Assignments submitted up to one week after the due date will get an automatic deduction of 10 points (i.e., if you submit your homework up to a week after the deadline, the maximum score you can receive is 90/100). Assignments submitted between one and two weeks after the deadline will get an automatic deduction of 20 points. Assignments submitted more than 2 weeks after the deadline will receive a score of 0 automatically. Please speak to the instructors if you believe you will need more than 3 weeks to complete an assignment.

The final grade will be based on homeworks (50%), a mid-term exam (20%) and a final project (30%). The midterm exam will be open book, open notes. Your class participation will also be considered in determining your final letter grade.

On-line Resources

Java 2 standard edition version 1.4 downloads
Sun's java documentation (jdk 1.4)
Companion web-site for the Java textbook. Has links to interactive quizzes for each chapter of the book
Corrections of errors in the textbook
Sun's java tutorial
On-line java tutor (assumes little or no programming background)
Usenet newsgroups:
- comp.lang.java.* (the * is a wild card that matches any characters. There are 16 different java related newsgroups under the comp.lang.java hierarchy)
- sci.med.informatics
  You can read these newsgroups through Google's free server or through a particular institution's server (e.g. news.mit.edu, news.dfci.harvard.edu, etc.)

Collaboration Policy

The instructors believe that collaboration is an important part of your educational experience, and you are encouraged to discuss the course assignments with your fellow students and to form study groups where appropriate.

To avoid crossing the line between collaboration and cheating/plagiarism, we require that you develop your own solution to each assignment. You may discuss strategies for solving programming assignments with your classmates and obtain assistance from them in debugging code that you have written. However, all code that you write and submit must be your own code with the following exception: When a homework assignment depends on code developed for a previous assignment, the instructors will provide solutions to the previous assignment that you may modify and extend to complete the new assignment. If it is determined that several people have cheated on an assignment, the grade for the assignment will be divided among all involved. For example if four people cheat and the work obtains a score of 100/100, each person will get an individual score of 25/100.

Please reference sources that contribute to your homework solutions.

No collaboration is permitted on exams.