Data Manipulation and Visualisation with Python
Monday 10 - Friday 14 July 2017
The King's Buildings, The University of Edinburgh, Edinburgh, Scotland, UK
Monday 26 June 2017 noon
Monday 3 July 2017 noon
15 (first come, first served)
first come, first served
£525 (includes coffee/tea, but no lunch)
One of the strengths of the Python language is the availability of mature, high-quality libraries for working with scientific data. Integration between the most popular libraries has lead to the concept of a "scientific Python stack": a collection of packages which are designed to work well together. In this workshop we will see how to leverage these libraries to efficiently work with and visualise large volumes of data.
"Martin was very good at both teaching and helping with customising/practicing on our own dataset." (September 2016)
Dr. Martin Jones (Founder, Python for Biologists)
Dr. Bert Overduin (Training and Outreach Bioinformatician, Edinburgh Genomics)
The workshop is delivered over ten half-day sessions. Each session consists of roughly a one hour lecture followed by two hours of practical exercises, with breaks at the organiser’s discretion. Each session uses examples and exercises that build on material from the previous one, so it’s important that students attend all sessions. The last two sessions will be kept free for students to work on their own datasets with the assistance of the instructor. A description of the sessions can be found at the bottom of this page.
Who should attend
This workshop is aimed at researchers and technical workers with a background in biology and a basic knowledge of Python (if you've taken the Introductory Python course then you have the Python knowledge; if you're not sure whether you know enough Python to benefit from this course then just drop Martin Jones or Bert Overduin an email).
Note that we are also offering a workshop Advanced Python for Biologists.
The Advanced course is aimed at people who want to develop bigger or more complicated programs in Python, or to learn more about the language, or to explore different approaches (object-oriented, functional) to programming. The material covered is very general purpose and can be applied to any kind of problem. The Dataviz course is about using a set of Python libraries that are specifically designed for data exploration and visualisation. In the Dataviz course, we're going to concentrate on using these tools to explore patterns in data, but the actual code that we write will be very simple - mostly we will be calling the functions and methods in these libraries.
To summarise: if you want to learn more about the language, build more complicated programs, or need to work with existing complicated programs, attend the Advanced course. If you want to explore datasets, find patterns, produce figures and charts, attend the Dataviz course.
Students should have enough biological/bioinformatics background to appreciate the example datasets. They should also have some basic Python experience (the Introduction to Python course will fulfill these requirements). Students should be familiar with the use of lists, loops, functions and conditions in Python and have written at least a few small programs from scratch. During the workshop students will use their own laptops. Instructions for any software to be installed will be sent out prior to the course.
1. Introduction and datasets
Jupyter (formerly iPython) is a programming environment that is rapidly becoming the de facto standard for scientific data analysis. In this session we'll learn why Jupyter is so useful, covering its ability to mix notes and code, to render inline plots, charts and tables, to use custom styles and to create polished web pages. We'll also take a look at the datasets that we'll be investigating during the course and discuss the different types of data we encounter in bioinformatics work.
2. Introduction to pandas
In this session we introduce the first part of the scientific Python stack: the pandas data manipulation package. We'll learn about Dataframes — the core data structure that much of the rest of the course will rely on — and how they allow us to quickly select, sort, filter and summarize large datasets. We'll also see how to extend existing Dataframes by writing functions to create new columns, as well as how to deal with common problems like missing or inconsistent values in datasets. We'll get our first look at data visualisation by using pandas' built in plotting ability to investigate basic properties of our datasets.
3. Grouping and pivoting with pandas
This session continues our look at pandas with advanced uses of Dataframes that allow us to answer more complicated questions. We'll look two very powerful tools: grouping, which allows us to aggregate information in datasets, and pivoting/stacking, which allows us to flexibly rearrange data (a key step in preparing datasets for visualisation). In this session we'll also go into more detail about pandas indexing system.
4. Advanced manipulation with pandas
In this final session on the pandas library we'll look at a few common types of data manipulation — binning data (very useful for working with time series), carrying out principal component analysis, and creating networks. We'll also cover some features of pandas designed for working with specific types of data like timestamps and ordered categories.
5. Introduction to seaborn
This session introduces the seaborn charting library by showing how we can use it to investigate relationships between different variables in our datasets. Initially we concentrate on showing distributions with histograms, scatter plots and regressions, as well as a few more exotic chart types like hexbins and KDE plots. We also cover heatmaps, in particular looking at how they lend themselves to displaying the type of aggregate data that we can generate with pandas.
6. Categories in seaborn
This session is devoted to seaborn's primary use case: visualising relationships across multiple categories in complex datasets. We see how we can use colour and shape to distinguish categories in single plots, and how these features work together with the pandas tools we have already seen to allow us to very quickly explore a dataset. We continue by using seaborn to build small multiple or facet plots, separating categories by rows and columns. Finally, we look at chart types that are designed to show distributions across categories: box and violin plots, and the more exotic swarm and strip plots.
7. Customisation with seaborn
For the final session on seaborn, we go over some common types of customisation that can be tricky. To achieve very fine control over the style and layout of our plots, we'll learn how to work directly with axes and chart objects to implement things like custom heatmap labels, log axis scales, and sorted categories.
In the final teaching session, we look at the library that both pandas and seaborn rely on for their charting tools: matplotlib. We'll see how by using matplotlib directly we can do things that would be impossible in pandas or seaborn, such as adding custom annotations to our charts. We'll also look at using matplotlib to build completely new, custom visualisation by combining primitive shapes.
9. / 10. Data workshop
The two sessions on the final day are set aside for a data workshop. Students can practice applying the tools they've learned to their own datasets with the help of an instructor, or continue to work on exercises from the previous day. There may also be time for some demonstrations of topics of particular interest, such as interactive visualisation tools and animations.