Overview of Requirements

• A paragraph abstract emailed to me by Monday, April 29, at noon. Your abstract should state what topic you would like to cover and what aspect of the topic you will go into detail on (see below).
• An (approximately) 5 page paper discussing your topic due the last day of final exams (May 18). Please email me a PDF of your paper as soon as it is complete.

This paper is to be done individually. But you are welcome and encouraged to discuss ideas with anyone in the class, and team up to and provide feedback on your reports. While the due date is late, you should treat this more as a 1-2 week assignment. The 5 page suggestion is assuming a generous double-spacing and margins. I provide that simply as a marker to ensure you do not write too much. I do not want you to spend more time than necessary and I also do not want to read 10 pages for 30 students. That being said, be sure to read the requirements below to understand what components you should place in your paper.

Introduction

Throughout the course, we we have seen several algorithms and structures for organizing and analyzing biological data. With each, you have built the foundations of your computational knowledge base that will aid you as you continue to mature as scientists. The constraints of a 14-week course limited the number of bioinformatics topics covered as well as the number of algorithms proposed to solve each problem. The central goal of this course, however, was not to memorize every bioinformatics algorithm. Rather, it was to develop the skills of algorithmic analysis and design that are a prerequisite to becoming a skilled computer scientist.

For your last assignment, you will write a medium-length paper reviewing a topic of research and study not discussed in class. There are dozens of potential areas of interest as well as various types of reports you can prepare. Below, I outline a few options as well as the expectations of your reports. You can pick from any topic related to the course (see below), including extensions to topics covered in class (e.g., probabilistic methods for tree inference). Your paper can have one of two styles: a review of a field of study or a review of a specific approach to a problem. I give examples of each and the expectations; in general, the former will require to discuss a little about many approaches/problems while the latter requires a more in-depth focus on one approach. Please be sure to email me a short preview of your topic by April 29.

Choosing a Topic

There are a wide-range of topics available to you. In general, it should be a topic related to bioinformatics, medical informatics, or machine learning. You may decide to cover something completely novel, or extend something covered in class. If delving into something we covered, you will be expected to go more in depth on the methodology as introducing the problem will be fairly basic. Some example biology problems:

• Genome assembly (taking many, small overlapping fragments and reconstructing the original genome)
• Single nucleotide polymorphisms (SNPs) and their uses (e.g., genome wide association studies (GWAS))
• Systems biology
• Network inference
• genomics and evolution
• Protein structure; protein function; secondary structure; protein disorder
• Mass-spec and proteomics
• RNA structure
• Protein-Protein interactions; protein-dna interactions
• Next generation sequencing data
• Medical informatics
• Translational bioinformatics
• Computational Immunology
• Image Analysis
• Databases and ontologies
• Biomedical text mining
• Disease models/ontologies
• Drug discovery
• Metabolic networks

If you are more interested in exploring general algorithms, you can use the following list to explore various techniques:

• Probabilistic graphical models (e.g., Bayesian networks, HMM's, conditional random fields, Markov random fields)
• Supervised learning algorithms (e.g., neural networks, support vector machines, deep learning)
• Unsupervised learning algorithms
• Sequential algorithms
• Vision/image analysis
• Memory efficient models
• Tree-learning algorithms
• Semi-supervised learning
• Search

Paper Options

You have two options for you paper:

A) Survey Paper

In this type of paper, you will review a general field of study. For example, you could motivate a problem such as RNA structure prediction and then briefly cover some of the proposed solutions (~1 paragraph each for 3 to 4 methods). Your paper will most likely rely on touching upon the findings of many papers or a central tutorial/review paper in the literature.

B) Method Review Paper

By this, I mean you are choosing to go in-depth in reviewing one specific approach to a particular problem. For example, I had you write a one page discussion on the T-Coffee approach to multiple sequence alignment. This type of paper will require you to concentrate on a detailed explanation of the technique, but you will need to only read one paper in-depth and possible another paper or two that provide further context.

Writeup Requirements

Your writeup should be well-structured and follow scientific writing principles. You can structure as you see fit, but at a minimum, a scientific paper touches upon these core topics:

• Introduction - Introduce the problem (computational and biological); preview contents of paper; this should sell me on reading the rest of your paper.
• Motivation - Is it clear what the purpose of the algorithm is? Why is the problem important to solve? Why is it difficult/not solved yet?
• Related work - How does the problem relate to a topic covered in class? How does it connect to other bioinformatics problems? How does it relate to non-bioinformatics topics (e.g., NLP, economics, robotics)?
• Methods - For paper A): what are the different types of approaches proposed to solve the problem? There may be dozens, but you should be able to describe the general theme of these methods (e.g., probabilistic; greedy; search-based). What are the most popular/common methods? Spend a paragraph each describing how 3-4 methods approach the problem.
  For paper B) you should go in depth on explaining the algorithm. You probably will want to write out pseudo-code and/or show an illustration of the algorithm overview. You should also have a paragraph outlining competing methods.
• Results and Experimental Methodology - How is/are the technique(s) evaluated? Is there a standardized data set that is publicly available? What are the measures used and what do they show?
• Discussions/Analysis/Conclusions - What are the advantages and disadvantages to the methods? Are there particularly types of data the approach works well on? What are the deficiencies? These should suggest future directions of research - what do you consider the most promising possible direction?

I do not expect you to report on every aspect of these topics, but touch on at least some of each core topic. Depending on your choice, it is okay to limit your research to one central paper or a light-reading of many papers. While this is hard to quantify, my expectation is that your paper will inform me about both your ability to frame the problem in a concise, yet informative and accurate manner. I also want to obtain an understanding of what you find interesting about the problem. Your paper should be typed; no hand-written writeups will be accepted. The exception is with illustrations, which can be neatly drawn and attached to your writeup. I would prefer you scan and attach the images to your PDF document, if possible. It may be useful to learn to use latex, a popular typesetting system that is used widely for writing scientific and mathematical papers. If you are interested, feel free to email me any questions about this. You can also find a primer and some sample guides online, including here or here.

Home Page