Before students began the online genomics CURE, formal written consent was obtained as part of a student experience and demographics survey. This project was conducted with an approved protocol through the Arizona State University Institutional Review Board (approval number STUDY00013025). Students were asked if they were at least 18 years old and if they consent to be part of this study. All students that consented to the study indicated that English was their primary language.

Students enrolled in this course were online students from a variety of biology majors (Biological/Biomedical Sciences, Neuroscience, Biochemistry, Conservation Biology, and Ecology), mostly in the final 2 years of their undergraduate degree program. Enrollees had interest in computational biology, but most had very little computer programming experience and many had not had opportunities to do research during their online degree program. To give students a basic foundation for computational analysis, students were required to take a prerequisite 7.5-week course in the first half of the semester (BIO 439: Computing for Research; session A in Fall 2022 and 2023). This course assumed no prior coding experience; students were introduced to command line programming, bash scripting, navigating a high-performance computing cluster, and ran through basic genomics analysis (fastq file quality control, alignment, and variant calling). The assignments in the prerequisite course are meant to be a broad first look at how students respond to computer programming and are not specific to the research project chosen for the CURE. All students were given the option of continuing on to the online genomics CURE (session B in Fall 2022 and 2023) but students needed to get approval from the lead instructor before being allowed to continue into the research session.

We have implemented 2 iterations of the online asynchronous genomics CURE covering 2 separate research questions (Fig 1, details in S1 Methods). The pilot iteration of the CURE was based on a bioinformatics project that investigated the effect of sequencing quality on inference of sex differences in gene expression in human placenta. Placenta gene expression data were processed with a range of parameters for sequence quality trimming software and students were asked to compare results to determine if the list of sex differentially expressed genes changed due to the stringency of the parameters chosen. The second iteration of CURE was based on inferring the sex chromosome complement in human cancer cell lines based on expression of genes on the sex chromosomes. Both of these projects were developed organically in the research laboratory of the course instructors, thus ensuring that the instructors had the required expertise to lead the course research and aiding the instructors in communicating why and how the research would contribute to the field.

While research training is typically conducted with a one-to-one mentor–student relationship, we employed a paradigm wherein a small team of instructors can provide mentorship to a larger group of students and simultaneously facilitate the development of a network of peers (Fig 1). During the first 2 modules, course materials and course communications were focused on getting all the students in the class the background information needed to understand the research aims and the way that the research is to be conducted. This included tutorials in the R programming environment, learning how the data were generated, and learning the biological, genomic, and statistical concepts necessary to understand the research aims. In Module 3, all students were assigned the same initial analysis by modifying template code provided by the instructors to facilitate troubleshooting coding errors and streamline guidance on how to interpret the results. For Module 4, students were split into smaller working groups to adapt the code to analyze different but related data sets that address the research question. Instructors assigned student working groups, putting communicative students that were showing technical skills with students that were struggling to maximize the likelihood that all parts of the divided work would be completed successfully and that students had the opportunity to learn from each other. Given small group assignments, students could have someone to work with if they desired or could just as easily work independently; students assigned to the same group could directly compare results and troubleshoot together if needed. Following Module 4, instructors checked data files generated by students for errors and put all the results from all student groups in a shared online location. For Module 5, students were asked to plot and interpret trends across the full data set. This module allowed struggling students time to catch up if they had coding issues in the previous modules and students that were ahead to expand on the analysis they had completed. In Module 6, each student was asked to present their results in the format of a scientific manuscript. Students were allowed to learn from the writing of the other students by conducting a peer review in Module 7. Students were also asked to turn in all code, figures, and output data files so that they can be used to prepare for publication after the completion of the course. In this way, we distributed the analysis, interpretation, and description of the research question equally among the students while simultaneously building in redundancy to make the research goals more robust to individual student challenges.

Learning objectives were written to match the knowledge and skills needed to perform each project within the seven-module format. The main objective for each iteration of the CURE was based on the research project chosen (Fig 1) and then mapped over 7 modules (listed in GitHub repository (https://github.com/SexChrLab/CURES)). Once the learning objectives were mapped out across the 7 modules, module learning pages, reading assignments, and coding assignments were developed to guide the students in achieving those learning objectives. Retention of the most important information and learning of skills in each module was tested using questions in learning assessments (quizzes) during each module. Project descriptions and matching learning objectives for both iterations of the course are available in our S1 Methods and GitHub repository (https://github.com/SexChrLab/CURES). Instructors analyzed weekly progress reports and assessments to address topics that needed clarification, common misconceptions, and ways to support student learning success in real time (Fig 2).

In lieu of expensive textbooks and akin to lab-based research projects, the instruction team wrote collaboratively to produce freely available reference materials to guide the students as they performed the analysis for the chosen research project. Each module had pages posted in the Canvas learning management system. In an introductory module (Module 0), students were introduced to the overall format of the course, introduced to the idea that they will be doing real research where they will collaborate to discover something new, and encouraged to engage in course communications. For Modules 1 to 7, learning pages were written in a modular fashion to keep the format of each module consistent so students could follow them more easily and to increase reusability for future CURE projects. Each learning page had 3 sections: (1) Biology/Statistics; (2) Coding; and (3) Professional Development. The Biology/Statistics section was used to describe biological relevance to the analysis, statistics necessary to test hypotheses being generated using the data available, and explain ways to interpret results. The Coding section featured formal instruction on computer programming and explained aspects of the code provided by the instructors. The Professional Development section showed the students ways to seek out information from the literature, described the publication and peer review process, and the cultural norms of scientific research. Additional learning resources for novices and enrichment for more advanced students were included as “Additional Resources.” Published bioinformatics workflows, vignettes, and tutorials were incorporated and/or modified as needed to demonstrate analysis relevant to answering the research question. Module learning pages for both iterations of the course are available in our GitHub repository (https://github.com/SexChrLab/CURES).

Writing learning pages for CUREs took less time with each iteration. For the first iteration, developing the course materials took about 1 to 2 weeks per module because the instructors developed all learning materials and assessments completely from scratch while simultaneously deciding on the overall framework of the course. However, developing materials for the second iteration was much quicker (about 3 days per module) because the overall framework for the course was already created and sections of the learning materials could be reused. The Professional Development sections were written to be evergreen and many of the Coding sections could be reused because both iterations of the CURE used code written in the R programming language. As the research being conducted in the CURE is based on the specialty of the instructors, we anticipate that instructors of genomics CUREs will build their own library of learning materials that can be reused, repurposed, or shared with other students for training outside the CURE from year to year.

To facilitate implementing functional code in a short timeline, the instruction team developed template code that the students were asked to modify and build on to complete the analysis required for the research project. Before the course started, instructors processed and prepared the starting input data. For Iteration 1 of the CURE, students were given template code to perform differential gene expression analysis that they could modify to run with different data sets. For Iteration 2, template code was provided to plot expression of a specific gene in cancer cell lines that they could modify to plot other genes and features of interest. Template code was provided in RMarkdown format so that code, products of code, and descriptions of code could be easily created in a single document used for grading, communication, and publication. Template code had many descriptive comments to explain each computational step and model best practices for coding technique. Tutorials assigned at the beginning of the CURE were chosen based on how well they explained the features in the template code so that students would feel more confident in making changes to suit the purposes of the research. Template code for each iteration of the course is provided in Github (https://github.com/SexChrLab/CURES).

As a final project in the CURE, students were asked to write a report in the format of a manuscript for publication and then asked to conduct peer reviews on manuscripts written by their classmates (Fig 2). This trained students to communicate scientific results with the level of accuracy and detail expected at a professional level. The professional development section of several modules was used to walk students through how to put together the various parts of their manuscript. Students were taught how to write detailed legends for figures produced as the research project was conducted in Modules 3, 4, and 5, then asked to create a storyboard outline of these figures to weave a complete set of related analyses that address the research project aims in Module 5, before fully writing about these results in Module 7. Examples of how to keep track of and document software packages used in code in the methods section of the manuscript were featured in all template code. Class discussions on topics involved in the research helped develop context that was featured in the abstract, introduction, and discussion portions of the manuscript. Many students wrote about results that were unexpected as those results fueled many discussions in lab meetings and Slack throughout the course, such as having a lot of overlap in results with different trimming parameters in Iteration 1 and cancer cell lines having sex chromosome gene expression that did not match what was expected based on the reported sex of the patient from which the cancer cell line was derived in Iteration 2. As all students did similar analyses, they were able to appreciate the descriptions and insights offered by their classmates during the peer review process. The rubric used to grade peer reviews of manuscripts is in the provided Github repository (https://github.com/SexChrLab/CURES).

The primary mode of communication with students was the Slack messaging system so that responses to questions and posts were visible to all students. This platform was integrated into the Arizona State University Canvas learning management system and allowed for asynchronous communication between students (see S1 Methods for details and alternatives). Students were encouraged to post and respond to Slack messages about confusing concepts, coding problems, and any other challenges encountered during the course. Instructors posted about common misconceptions, bug fixes, activities and resources for learning enrichment, and encouragement for students as they struggle with challenging material. In addition to Slack, students were given several opportunities each week to meet with the instruction team synchronously: lab meeting, writing hours, and shared research hours (Fig 2). Optional weekly lab meetings recorded and transcribed in Zoom covered student-reported challenge topics. Optional weekly writing hours gave students the opportunity to discuss interpretation of findings they would write about in their weekly assignments and manuscript. Optional shared research hours were offered twice a week, giving students the opportunity to troubleshoot code with the instruction team live with options to share their screen.

In the second iteration of the CURE, instructors implemented an asynchronous journal club to help students engage with peer-reviewed literature relevant to the research project (Fig 2). To do this, the instructors chose 1 publication each week for the first 5 modules for the students to read (leaving the last 2 modules with more time to focus on the manuscript and peer review). To encourage collaborative learning, the chosen publications were posted using the Perusall social annotation environment and students were given points toward their final grade for posting comments as they read the paper (see S1 Methods for more information). We found that the students pointed out what they thought was interesting or relevant about the publications and asked and answered each other’s questions about the journal articles. This format of journal club provided the flexibility needed for online student learning while providing asynchronous but meaningful discussions about important published works in the field.

Custom rubrics and examples were provided to students to describe expectations for assignments and scientific writing. The majority of the course grade for the CURE was based on the weekly progress report so the majority of the grading efforts of the instructors were dedicated to this (Fig 2). As the progress report was divided into sections, instructors were able to split the work, typically with one instructor grading the accomplishments and challenges sections and another grading the scientific writing and coding assignments addressing the research aims. Instructors held a short weekly meeting to discuss research progress and how to address student challenges. Since a great deal of the grade for the CURE was based on open-response writing, instructors watched carefully for signs of academic dishonesty, including inappropriate use of artificial intelligence chatbot engines and copying between students (S1 Methods), but because so much emphasis was placed on progress over products, the vast majority of students submitted their own work and experiences throughout the CURE.