Online Tools for Teaching Cancer Bioinformatics

ABSTRACT

The rise of deep molecular characterization with omics data as a standard in biological sciences has highlighted a need for expanded instruction in bioinformatics curricula. Many large biology data sets are publicly available and offer an incredible opportunity for educators to help students explore biological phenomena with computational tools, including data manipulation, visualization, and statistical assessment. However, logistical barriers to data access and integration often complicate their use in undergraduate education. Here, we present a cancer bioinformatics module that is designed to overcome these barriers through six exercises containing authentic, biologically motivated computational exercises that demonstrate how modern omics data are used in precision oncology. Upper-division undergraduate students develop advanced Python programming and data analysis skills with real-world oncology data which integrates proteomics and genomics. The module is publicly available and open source at https://paynelab.github.io/biograder/bio462. These hands-on activities include explanatory text, code demonstrations, and practice problems and are ready to implement in bioinformatics courses.

INTRODUCTION

As data become a dominant feature of biology (1), university curricula must adapt and prepare their students for a data-centric future (2). Many life science undergraduates do not receive bioinformatics training, and a majority of professional biologists gain bioinformatics skills without institutional training (3). This is often because scientific computing courses are designed for computer science and engineering majors (4). Beyond simple introductory courses, there is a need to develop advanced curricula which truly integrate computational and biological perspectives (reviewed in reference 5) as either life science electives or a formal bioinformatics degree (6, 7). A key design strategy is identifying authentic, biologically motivated computational courses, including hands-on exercises with real data (8).

As an emerging discipline, bioinformatics is evolving rapidly. Early coursework and textbooks for bioinformatics focused on algorithms, e.g., Rosalind (http://rosalind.info/). However, the rise of deep genomic and proteomic profiles has highlighted a need for instruction in exploratory data science and statistics. Massive public data sets offer an incredible opportunity to integrate molecular and cellular biology with computation (9), and published classroom activities exist for a variety of topics, including DNA assembly (10), metagenomics (11), bacteriophage genomics (12), and viral genomics (13). Unfortunately, current topics largely focus on genomics and not a broader genotype-to-phenotype view of biology. Additionally, many public data sets have practical barriers that complicate classroom use, such as data accessibility and package installation (14). Here, we present a cancer bioinformatics module with six exercises to teach an integrated proteogenomic view of cancer. The exercises contain explanatory text, code demonstrations, and practice problems encoded in publicly available Google Colab notebooks intended for online or traditional instruction.

PROCEDURE

Computational cancer biology module

Improvements in cancer diagnosis and treatment will come from analyzing DNA, RNA, and protein data, along with traditional clinical information (15). We present a module for teaching computational cancer biology to demonstrate how modern omics data are used in precision oncology. Each lesson explores fundamental cancer concepts and supporting molecular data (Table 1). In lesson one, students are introduced to the proteogenomic data set (16–19). As DNA mutation is the root cause of cancer, lessons two through four describe three distinct kinds of mutations and their impacts on the genome. Lessons five and six use transcriptomics and proteomics data to identify differential gene expression and activated pathways. Each lesson also teaches computational tools and analytical techniques essential for modern cancer research. The first lesson introduces pandas DataFrames and manipulation of data matrices. Subsequent lessons expand the students’ skill set with DataFrames, application programming interfaces (APIs), statistical methods, graphing, and network exploration.

TABLE 1.

Six lessons and associated learning outcomes: at the end of each lesson, students will be able to perform the following learning outcomes

Lesson topic	Biological learning outcomes	Computational learning outcomes
1. Introduction to Cancer Datasets	• Locate proteogenomic cancer datasets from real tumor samples	• Manipulate Pandas DataFrame
		• Select relevant clinical cancer data
2. Missense Mutation	• Identify frequent mutations in cancer cohorts	• Access UniProt knowledgebase using API
	• Compare and contrast different cancer types	• Parse JSON and integrate cancer data with UniProt
	• Assess functional impact of DNA mutation
3. Truncation Mutation	• Classify mutation impact on primary sequence, domain structure, and protein function	• Create new DataFrame elements with .apply() function
		• Access UniProt knowledgebase using API
		• Import/export data from Colabs
4. Copy Number Variation	• Identify copy number variation (CNV) changes in genes	• Plot CNV events
	• Distinguish between focal and arm level CNV events	• Integrate gene location and CNV data
	• Compare frequent CNVs in multiple cancer types	• Create new DataFrame elements with .apply() function
5. Transcriptomics	• Identify differential expression of transcripts between tumor and normal samples	• Perform pathway enrichment analysis
	• Interpret a gene list as a functional set of pathways	• Create boxplot visualization
	• Justify choice of enrichment methods based on gene list characteristics	• Perform unpaired t test with multiple hypothesis correction
6. Proteomics	• Evaluate utility of protein coexpression networks based on agreement with known protein interaction networks	• Perform correlation analysis
		• Create network visualization
		• Access UniProt knowledgebase using API

Overcoming instructional barriers

The module is designed to overcome both practical and philosophical barriers to student engagement with the integration of three computational tools. First, lessons are coded in Google Colab, a live Python coding environment accessible via a Web browser (https://colab.research.google.com/). Importantly, Colab does not require any software installation and works with all operating systems, including mobile platforms. Thus, students have access to a functional computing environment on the first day of class, without requiring information technology (IT) support. Second, real cancer data sets are streamed via the cptac Python API (20), providing access to genomic, transcriptomic, proteomic, and clinical data. Mundane and frustrating tasks like locating files or loading data are automated, thus saving time for students to focus on cancer-relevant computation. Third, our module uses an autograder, which contains homework answers and hints used to give students immediate feedback and provide help on integrated practice problems.

Student preparation

The module is designed for upper-division undergraduates. For biological preparation, we suggest that students have completed courses in molecular biology and genetics; for computational preparation, we suggest that students have completed courses in Python programming and introductory statistics. The first lesson, called Introduction to Cancer Datasets, is a skills check. A prepared student should be able to complete it in about 30 to 45 min. This lesson is the same as the final homework in the author’s (S.H.P.) Introduction to Bioinformatics course, and students with one semester of programming would be prepared computationally. If students have difficulty with this first assignment, then it is not advised to continue with the module.

Using the module in the classroom

We created this module for those wishing to teach computational cancer biology or expand their offering of bioinformatics topics to undergraduate life science majors. The module and links to the six lessons are available at https://paynelab.github.io/biograder/bio462. Each lesson (except for Lesson 1: Introduction to Cancer Datasets) is designed to contain a week’s worth of instruction material, with approximately 6 h of computational exercises. Although each lesson is independent, the later lessons build on the computational skills from earlier lessons. Therefore, the module is intended to be used as a set.

Lessons are delivered through Google Colab notebooks and interweave descriptive text and software code. Notebooks begin with a section describing the topic and relevant literature, which could be used in classroom lectures or preparatory reading. The explanatory text motivates students with the description of a cancer biology problem and then proceeds with demonstrative software examples that teach students core computing concepts. These are followed by practice problems where students write their own code. Each practice problem is graded with the autograder, and students can ask for hints on how to approach the question. As with any course, it is advised that a teaching assistant with a bioinformatics background be available to guide students through the rigorous lessons. Instructors interested in the answer key may contact the authors.

Student feedback

Student evaluation of the lessons and module contained both supportive and critical feedback. In general, students were enthusiastic about the authentic integration of computing and cancer biology. A significant positive of the lessons was the links to articles that prompted a deeper exploration of biological concepts and a reference for algorithm use or syntax. The most common challenge was learning to work with data matrices; for students who learned programming through the object-oriented lens, computational exploration of DataFrames was an adjustment. Student feedback on specific exercises prompted substantial revision in wording, hints, and primer examples.

Safety issues

This module is purely computational and does not include any laboratory components. The data used in the modules are publicly available and do not include any personally identifiable information.

Conclusion

Cancer bioinformatics is a growing area, and proteogenomic analyses of large cancer cohorts are rapidly changing cancer diagnosis and treatment. Many of these rich data sets are publicly available, enabling the creation of engaging and impactful curricula. We created a module that introduces undergraduate students to proteogenomic cancer data and teaches bioinformatics skills in data analysis and interpretation. By utilizing a unique combination of technologies, the module avoids common obstacles to large-scale and authentic bioinformatics exercises in a classroom setting.