TL;DR
This paper describes a successful undergraduate and graduate-level programming course designed to bridge the gap between computer science and molecular biology. The course equips students with essential programming skills and foundational knowledge in bioinformatics, enabling them to contribute effectively to computational biology research.
What Problem Does This Paper Address?
The authors identified a growing need for individuals skilled in both computer science and molecular biology, particularly for research assistant roles. They observed that computer science students often enter biological research without a strong understanding of the biological "big picture" or the unique challenges of biological data (e.g., noise, complexity). Conversely, biological sciences students might lack the computational expertise to leverage modern tools. This gap leads to communication issues, inefficient problem-solving, and a tendency for technically-minded students to apply algorithms to overly familiar, less impactful problems. The paper addresses this by proposing and evaluating a structured programming course specifically tailored for non-biology students interested in computational molecular biology.
Methods
The core of the paper is the description of a programming-intensive course titled "Representations and Algorithms for Computational Molecular Biology." The course was structured around three major projects and six smaller assignments, designed to teach fundamental representations and algorithms in computational molecular biology.
Key pedagogical approaches included:
- "Just Enough" Biology: Biology was introduced in a "skewed" manner, focusing on concepts directly relevant to the computational tasks. For example, the functional characteristics of globin molecules were explained to motivate sequence alignment and structure prediction.
- Hands-on Programming: The emphasis was on students building tools rather than just using existing software. Students were allowed to use any programming language, and assignments progressively required more significant programming.
- Project-Based Learning: The course featured substantial projects that taught students how fundamental algorithms work, where they might fail, and their sensitivity to data errors.
- Integrated Lectures: Lectures covered core concepts and then delved into faculty research interests, providing real-world context.
- Jargon Management: Instructors actively monitored for and clarified biological jargon, and a "virtual email server" was provided for quick definitions.
Datasets
The course utilized publicly available biological data, including:
- DNA and Protein Sequences: Retrieved from GenBank.
- Protein Structures: Accessed from the Protein Data Bank (PDB).
- Specific Protein Examples: Sperm whale myoglobin and hemoglobin were used extensively as case studies for sequence and structure analysis. Hemoglobin and Superoxide Dismutase were used for 3D structure analysis.
The paper does not introduce new benchmark datasets but rather leverages existing, well-established biological data resources.
Key Insights and Ideas
- Bridging the Disciplines is Crucial: The course demonstrated a strong student interest in computational biology, highlighting the demand for interdisciplinary training.
- Programming is Key: Students appreciated the "hands-on" programming aspect, reinforcing the idea that computer science departments should equip students with prototyping environments.
- Biology as Motivation, Not Barrier: Teaching "just enough" biology proved effective in motivating computational tasks without overwhelming students with biological detail.
- Structure Informs Sequence Analysis: The course's emphasis on structural molecular biology helped motivate critical issues in sequence analysis, challenging the traditional view of starting with string matching analogies.
- The Need for a Core Textbook: The authors noted the difficulty in defining a universal core curriculum for computational molecular biology courses, underscoring the need for a high-quality, general textbook.
Key Results
The course was well-received and considered a success.
- High Enrollment: The course exceeded its target enrollment, attracting 50 attendees initially and ultimately enrolling 24 students for credit, a mix of graduate and undergraduate students from Computer Science, Medical Information Science, and Engineering.
- Positive Student Feedback: Students appreciated the hands-on interaction with data and programming.
- Core Requirement: The course was made a core requirement in the Medical Information Sciences program.
- Student Engagement: The course led some students to pursue computational molecular biology efforts.
- Successful Culmination: A final project involving protein threading to identify globins was a successful challenge for the students.
Limitations and Future Work
- Time Constraints: The 10-week quarter format limited the depth and breadth of topics that could be covered in assignments. Specific areas like secondary structure prediction and fragment reassembly were covered in lectures but not in assignments.
- Reading Load: Students struggled to complete assigned readings due to the heavy programming load.
- Future Work:
- Spend more time on dynamic programming, including variations and scoring matrices.
- Moderately reduce emphasis on structural computation in assignments, shifting towards sequence statistics for classification.
- Provide more biological background material throughout the course, rather than concentrated at the beginning.
- Reduce the amount of readings and integrate them more closely with assignments.
- Investigate ways to create a more readable syllabus of papers, acknowledging the need for a high-quality textbook.
Summary updated