Week 15
Final project presentations
The final project, presented this week. It is 45% of the grade: proposal 10%, progress update 10%, final report and presentation 25%. A good project picks a problem where an application challenge and an algorithmic component meet. Either benchmark established methods (at least one new to the application) and explain why they succeed or fail, or change one component of a method based on a hypothesis about the application and test whether it helps. Teams have up to three members, and expectations rise with team size. The proposal and update are early drafts of the report's introduction and methods and results.
Objectives
- Choose a project topic that pairs an application (structure, function, interaction, or mutational effect prediction; protein or drug design; omics or systems analysis) with an algorithm or data modality, and state the challenge the algorithm is meant to address.
- State a testable hypothesis of the form 'component X addresses challenge Y in application B', rather than 'algorithm A works for application B'.
- Write a two-page proposal covering the problem, its significance, current methods and their gaps, the proposed method and rationale, data, assessment plan, and timeline with milestones.
- Design an evaluation: a dataset from a published paper (ideally with code), train and test splits that keep similar sequences apart, baselines, and metrics suited to the task.
- Run the comparison or ablation that tests the hypothesis, and interpret the result either way: an improvement supports it, and no improvement points to the component, confounders, or a next variant to try.
- Write a two-page progress update: progress, obstacles and new ideas, current results, expected outcomes, and an updated timeline.
- Give a 15-minute presentation, including 2 minutes of questions, covering the problem and its challenges, the method and its rationale, and actual or expected results with insights and conclusions.
- Write a 6–8 page final report in paper format that could be extended into a workshop submission.