Writing Bioinformatics Methods

Important

Bioinformatics methods sections should convey the same level of information as a wet-lab methods section. They should be:

• clear
• reproducible
• adequately cite the tools used

Caution

Be careful if using a published paper as a template for your own methods section. Many published papers do not include sufficient information in their bioinformatics sections to be reproducible, and make for bad examples.

• Annotated example of how to write an experimental methods section - An example of a generally good experimental methods section and some key points where it might be improved - LP
• Annotated example of how to write a bioinformatics methods section - An example of a generally good bioinformatics methods section and some key points where it might be improved - LP

1 Tips for Writing a Great Methods Section

• Describe your experimental methods in enough detail that your reader could replicate the experiments.
• Your methods section is not a protocol (a step-by-step list of everything that was done.) You may assume that your reader is a competent scientist with the skills to perform basic lab calculations, techniques, etc.
• Cite the appropriate literature (i.e. for any recipes, tools including software, and protocols you have used).
• Do not include results or rationale for performing experiments (these belong elsewhere in your thesis).
• Be clear and concise; eliminate unnecessary words and steps.
• Make sure that your methods section includes all of the materials and methods you used in your thesis (This includes any bioinformatics methods such as BLAST!)
• Ask a friend or colleague to read your methods section and give feedback.

Tip

Look at papers that have used similar methods: how are these methods sections written? Though take care as many papers underdescribe bioinformatics methods.

You want to make sure that your methods section generally follows the conventions in your field, so that readers will be able to understand it more easily.

A few examples of specific conventions:

• for wet-lab projects, centrifuge speeds are always given in xg, not in rpm or rcf
• for science communication projects, the design of your intervention/survey belongs in the methods section
• for bioinformatics projects, always include the version number and citation for any software used and details of parameters, etc.

2 What I look for when reviewing bioinformatics methods

Overall, I’m looking to understand two things:

1. Where the data comes from
2. How the data was processed

2.1 Where the data comes from

The kind of information I need to understand whether the analysis has a sound set of inputs includes:

• what experimental method(s) generated the data (if any)?
• what database were the data obtained from
  • as databases and their contents can change over time, this might require a version number or date of access
  • how were the data identified in the database? Was this a search result (if so, what search terms were used)? If not a search result, how was the data reduced to the subset you’re working with?
• were the data processed after you obtained it?
  • this might include filtering and QC for sequence reads, for example - or excluding sequences above or below a particular length

2.2 How the data was processed

There are two overarching components of the research I want to know from your methods section:

1. How was each step of analysis or data processing performed?
2. How do the individual analyses flow together to take the data from its initial form to a final output?

A flowchart figure (as in the example linked above) can be a very simple and effective way to deliver the necessary information for part (2). This doesn’t need to have every last bit of detail, but does need to get across to me how the data was taken through each step, in what order, and what general state the data was in between each step.

Especially for involved, complex analyses, the information required to satisfy part (2) can be quite bulky. For any given step in the analysis I want to know:

• what was the input data?
• what tool was used (e.g. a software package or a script, with an identifiable name)
• what version of the tool was used (e.g. software version number, or link directly to the script on GitHub)
• a citation for the software, if it has been published
• the parameters used for the analysis

There is quite a bit of flexibility, here. It is certainly possible to detail every piece of information in the Methods section of the thesis (or paper), but it is also permissible to put much of the detail into supplementary information or an appendix, or make it available online, so long as the reader can understand what was done and the information is available to them.

For example, if you performed each individual step manually at the command-line, it would be reasonable to report each command-line in the Methods section.

Alternatively, you might have automated the analysis into a script, or Galaxy workflow, in which case you could make the script/workflow available online, and put the link into the Methods section. This would mean you only need to outline the analysis process (with tool names, versions, citations, and relevant choices), but could leave the detail of the commands to interested readers who followed the link to the script.