Breaking Barriers in Large Classes: Using Cloud Computing and Computational Biology to Implement Research‐Inspired Curricula in Biochemistry I

ABSTRACT

Course‐based undergraduate research experiences (CUREs) are a high impact instructional approach but are challenging to implement in large lecture courses due to class size and content demands. We describe the evolution of a research‐inspired project in a large‐enrollment (70+ students) Biochemistry lecture course. Our goal was to integrate a project that complements the rigid content while providing experience working on a novel research project. We addressed access barriers by expanding the use of cloud‐based computational techniques, notably Google Colaboratory (Colab), to ensure software accessibility across diverse student devices. The semester‐long project is divided into four modules focusing on protein structure and function, culminating in an individual project that studies a novel protein variant and its effects on structure and function. These projects are further supported by a science and math librarian to teach information literacy and data management skills. Since Fall 2024, over 250 students have participated. Student self‐assessments demonstrate gains in confidence, particularly in analyzing molecular structures, connecting genetic mutations to protein function, and using scientific databases like UniProt. This model shows that research‐inspired curricula can be successfully integrated into large, content‐heavy lecture courses by leveraging accessible cloud‐computing resources.

1. Introduction

Course‐based undergraduate research experiences (CUREs) and related course‐based, research‐inspired projects have proliferated across instructional laboratories nationwide [1, 2, 3]. In these course formats, students undertake a novel research project to gain experience asking and answering scientific questions, as well as with lab techniques [3]. The scientific questions are frequently novel, but the level of student involvement in researching and designing the question can range from research directions designed by the instructor to entirely student‐designed and driven by the instructor [4, 5, 6]. Data from student learning assessments suggest that students make gains in several areas, including confidence, identity as a scientist, and data interpretation [7, 8, 9, 10]. These student gains and the benefits to faculty scholarship make this a popular teaching approach.

CUREs and research‐inspired curricula have faced barriers to being used in lecture settings, including large class sizes, access to viable projects, and conducting experiments in a non‐laboratory setting. Laboratory courses are the traditional realm for applying knowledge, while the related lecture course is for learning content knowledge. Previously, we described a hybrid lab‐lecture approach that incorporated bioinformatics approaches in a Biochemistry II lecture to complement laboratory experiments [11]. While the smaller course environment is simpler, the various benefits of learning using a research project are restricted to a generally enthusiastic and self‐selecting population. Moreover, not all institutions offer an advanced biochemistry course. Thus, our goal was to expand this class format into a more content‐heavy course with a larger class size, providing as many students as feasible with the opportunity to participate in research‐inspired coursework.

One way to do experiments and research in a lecture setting is to use computational techniques. The use of computational research techniques in CUREs has increased, especially in response to the COVID‐19 pandemic and the shift to online learning [12, 13, 14, 15]. The implementation of computational work is restricted by the computational capabilities of the student population and the devices that they bring to the classroom. However, the increased availability of affordable cloud computing and web‐based coding interfaces has made computational experiments more accessible to undergraduate scientists with varying levels of computer experience and device capabilities [16, 17, 18, 19]. In addition to existing servers, there has been an expansion of platforms that run user code in the cloud and allow for the sharing of existing code that others can run, such as Google Colaboratory [20]. These recent gains in computational biology and the accessibility of research‐grade modeling software present an opportunity to use these resources to conduct research projects in a lecture setting and at a larger scale than before. Beyond accessibility issues, Biochemistry I is often the final chemistry class for students and covers a wide range of topics, spanning general chemistry to molecular biology, and is often subject to the tyranny of content deemed “required” to prepare students for everything from medical school/MCATs to graduate work and the workforce. The topics in a “standard” Biochemistry I course include biomolecular structure, enzyme function, and an introduction to metabolism. Thus, a research project that is incorporated into Biochemistry lecture should be designed to support, or at least not interfere with, the content learning goals.

Since our 2016 publication, we have revised the lecture‐based project from a class of 10 in a seminar‐style course to a one‐semester Biochemistry I lecture class of 70+ students. As with any project, numerous challenges arose as the scale and environment changed. These issues include engaging a more diverse audience in the research process, the variety of student devices, the increased use of mobile devices rather than laptops, and managing a large‐scale, course‐based research project alongside the more rigid curriculum expected in a Biochemistry I course. This manuscript builds on our earlier work to describe the evolution of the project to a larger class setting designed for a mix of student majors and professional goals. The research project does not decrease the content covered in the course but rather complements and centers the content, allowing students to apply the knowledge from the textbook. During development, we expanded the range of computational techniques and incorporated browser‐ or cloud‐based resources to broaden software availability across a wide range of student devices. Finally, we describe two more recent developments that enable the rapid dissemination of student projects and permit students to access cloud‐based high‐performance computing.

2. Methods

2.1. Assessment

Students were asked to complete a survey before the beginning of the projects and at the end of the semester. The survey is a self‐assessment of their confidence in the skills used in the course. Students ranked their confidence using the scale from 1: Could perform the task with step‐by‐step guidance from the instructor to 5: Could perform the task with almost no supervision or guidance. The data were anonymized, aggregated, and analyzed using R. The James Madison University (JMU) Institutional Review Board (IRB) monitored the activities and assessment of this work as outlined in protocol 21‐1961 and classified it as exempt.

3. Results

3.1. Course Environment

We have deployed the research project format into two biochemistry classes at James Madison University, a mid‐sized R2 university. The first course setting is a 300‐level course for chemistry majors and non‐majors (premeds, biology, etc.), which covers atoms to metabolism in a single 14‐week semester. The textbook used in class was Biochemistry: Free for All [21], which is a free open educational resource (OER), and we supplemented this text with sections from “Structure and Mechanism in Protein Science” by Alan Fersht, which is also freely available online [22]. The second course in which we have incorporated the research‐project curriculum is a 200‐level Biochemistry course for the Dietetics major at JMU. This course is also popular among pre‐PA students who are typically biology or health sciences majors. Roughly 70 students take this class in a single session, and a laboratory course is a co‐requisite, which is normally taught in part by the lecture instructor. The same textbooks are used as in the 300‐level course; however, the material is refocused to include more links to human health, and the content on biochemical techniques is shifted to the laboratory course, allowing more time to cover metabolic topics. Both lecture classes are taught by a single instructor, with 0–2 undergraduate learning assistants who provide in‐class instructional support and hold ~3–4 h of study (office) hours per week, in addition to the instructor‐hosted study hours but do not grade assignments.

In both classes, students learn content knowledge outside class and are assessed through quizzes hosted by the course LMS, with class time devoted to project work on problem sets or case studies to help them understand the material. Both lecture classes meet two times a week for 75 min. Both the instructor and undergraduate Learning Assistants (LAs) hold study hours during the week, and students can receive points, called Weak Interaction points, from attending these sessions. Students are not required to attend these sessions and there are other mechanisms to earn these points, including supplemental assignments. About 90% of students attend at least one study hour during the semester. Further support for topics in literature searching and data management is provided by the JMU Libraries Science & Math Librarian. The librarian presents approaches to searching the scientific literature during 1–3 class sessions and is available for office hours or consultations.

3.2. Project Redesign and Organization

Early on in the project redesign and expansion, three key problems appeared: (1) a minority of the students had experience searching the literature and scientific databases to contextualize their work, (2) not all students had devices capable of running PyMOL or other software tools that are common in science, and (3) it was difficult connecting the projects to the “standard” curriculum found in a Biochemistry I course serving a wide range of majors and outcomes.

As we addressed connecting the research goal to the course curriculum, we found that dividing the project into four smaller projects which we directly connect to broader themes in the course like biomolecule function. A brief description of the smaller project goals for the Fall 2024 to Fall 2025 iteration, along with the timing relative to the course content topics, is shown in Table 1. The first project has students construct a brief biography of a biochemist. It is aimed at developing the student team in a relatively low‐stakes environment and at strengthening students' ability to search the literature. Project 2 is focused on having students predict the structure of a protein and assess whether the molecule is folded well using their knowledge of biomolecule folding complemented by software such as MolProbity. The third project has students explore the function of the protein and ultimately propose a catalytic mechanism. The fourth and final project asked students to do a task incorporating the skills from the previous projects. Typically, this assignment asks students to detail how a variation in the protein structure due to a mutation or post‐translational modification can change the function of the protein and lead to metabolic or physiological changes. Examples of the project handouts are provided as Supporting Informations. During the first 8 weeks of the semester, students work in teams of 3–5 members on the first three projects. For the fourth and final project, students work individually, and the final product is due on the last day of the semester giving students roughly 6 weeks to complete this assignment.

TABLE 1.

Alignment of project and content knowledge.

Project name	Project goal (Fall 24 to Fall 25)	Software (Software offered via Colabs indicated with *)	Course topics covered during the project
Awesome biochemists	Learn to search the literature, team building	Zotero/Zbib [23]	General chemistry review, Intro to biomolecules
Homology modeling (Structure)	Make a model of a MDH1/2 and assess the structure quality	AlphaFold3 [24], Intellifold [25], Molprobity [26], Molstar [27], MD [16, 28]/NMA* [29], Analysis of MD/NMA* [30]	Protein structure, biochemical/structural techniques
Informatics and Docking (Function)	Draw a reasonable enzyme catalytic mechanism in the active site and how the substrates of MDH1/2 bind	Molstar, Uniprot, PrankWeb [31], AlphaFill [32], Intellifold, ProLif [33, 34], PropKa* [35], AlphaMissense [36]	Enzyme mechanism and kinetics
Studying/designing a variant → effects on broader metabolism	Develop a hypothesis based on modeling data describing how novel amino acid variation in MDH1/2 leads to functional and structural changes in the protein	AlphaFold3, Intellifold, MD/NMA*, MD analysis*, Molstar, PropKa*, AlphaMissense	Central dogma and metabolism

For each project, 1–2 class periods are dedicated to getting teams started with the software and addressing any questions they may have. Between these project sessions and the librarian sessions, more than a quarter of the 26 class periods in the semester became occupied by the project or skill development. During non‐project sessions, students carry out in‐class activities, such as those described by us previously on enzyme kinetics [37], or work through case studies or similar activities, such as effects of sequence variation on hemoglobin structure/function, which reinforce the content areas. This delivery style also gave students some freedom to bring project issues on non‐project days to receive help. Content knowledge was assessed through out‐of‐class problem sets hosted in the course learning management system, which were based on the textbook readings, with connections to project content or prompts that used project skills. Project deadlines were set so that the end of a project coincided with a move to a new theme within the content, such as biomolecule structure, enzyme function, or metabolic function and disease. Thus, the projects were woven into the course structure as a learning tool rather than replacing the content delivery and expanse.

To balance project work and content knowledge, each project assessment has both a team component and an individual component. The team component is the biography or scientific results of their modeling. The individual assessment includes a reflection on team cohesion, a narrative on what the student did not understand during the project, and 1–5 questions or tasks assessing the student's understanding of how the content material could be applied to the project topic. These latter questions include asking students to design inhibitors based on their team's proposed enzyme mechanism or to show examples of different kinds of weak interactions and structural elements within the protein that their team modeled. Project 4 is an individual task and serves as the “final exam” for the course, accounting for approximately 15%–20% of the final grade. Students are assessed on their ability to complete tasks (such as modeling the variant structure) and their ability to provide reasonable data supporting their proposal on how the variant affects structure and function (or doesn't). The data and proposal are compiled in a final paper, and in the recent project, the students further summarize their findings in a README on GitHub, along with the key data. This README file renders as a web page with figures and text and is a publicly available version of their project. During the Fall 2024 to Fall 2025 iterations of the courses, students were tasked with describing how phosphorylation affected the structure and function of human MDH1 or MDH2, and each student was assigned one of the ~100 known and potential modification sites to study which has been deposited on GitHub (https://github.com/CEBerndsen/MCC_MDH_DB).

3.3. Incorporation of Web‐Based Software and Databases Into the Projects

The easiest project to revise was the second project on protein structure due to the numerous structure prediction and characterization servers hosted by EMBL and others [17, 24, 25, 26, 27, 32, 36, 38, 39] and our previous focus in this area [11]. Projects 1, 3, and 4 were built on this initial project, with the literature search in project one coinciding with a librarian collaboration, and projects 3 and 4 emerging from the availability of biochemical software and the use of Colabs. A second aspect of the project redesign was to narrow the scope of projects 2 and 3 to ensure uniformity within the class. These early projects focus less on generating novel results and more on skill development, using the software, and learning to appropriately visualize biomolecules within a project where the outcome is reasonably predictable to the instructor. The final project opens students to a novel direction for their project, built on their early work. Topics have included the effects of variants and the design of enzyme inhibitors, which align more closely with well‐known CURE or research‐inspired project formats. Effectively, the data from earlier projects serve as the control data for students to determine whether variation affects protein structure or function. Students must individually repeat many of the tasks from earlier projects, including finding information on known diseases or phenotypic links to the variation, modeling the structure and dynamics, and determining potential changes to substrate/ligand binding. Project outlines and the questions from the individual parts of the projects are in File S1.

3.4. Incorporation of the Library Activities

Another early curricular modification was the addition of a librarian to the instructional team. The conception and impetus for the librarian visits have been widely covered in the literature [40, 41, 42, 43], and we have previously described a version of this collaboration [44, 45]. The librarian is a key resource for students, supporting both literature research and data management aspects of the projects. Across both courses, the librarian provided core information literacy concepts related to finding, using, and evaluating scientific sources [46]. The librarian typically taught 1–3 sessions across the semester, and each in‐class workshop employed a combination of the following: using research maps to build research skills and background knowledge [47]; using known literature sources to “search in time,” or the metadata basics needed to organize research data generally, and specifically in GitHub. These topics were strategically aligned with the course's project requirements.

The librarian's first session is timed to align with project 2. Because students come to the research process with various skill levels, a research mapping exercise was used to normalize the research process and gain familiarity with library resources across learners. The mapping exercise asks students to build their background knowledge, to make predictions, and to engage with library databases. The mapping exercise uses reflective learning techniques, such as drawing and brainstorming, to encourage students to engage with the more tacit components of research. Students often believe that there is “one right answer,” so this activity promotes open inquiry, which is required for making predictions throughout the course. The second session aligns with project 3. At this juncture, students are beginning to dive deeper into the scholarly literature. The “searching in time” lesson shows how to use known literature sources to search forward or backward in time, using an article's bibliography or “cited by” features in databases. This strategy is conveyed as a way to pause and reflect when the research feels overwhelming, or when students find that their “research rabbit holes” lack direction. As project 4 begins, students receive top‐level instruction on the history and role of metadata in everyday life, and the importance of data management, personally and professionally. Additionally, students receive detailed instructions on how the metadata in their README markup files will be used in GitHub. Markup files are often a new concept for students, so the librarian details how good project metadata will be transformed into a webpage that other researchers can use. These lessons and their associated materials are available for reuse and remixing from the linked OSF repository and are provided in the Supporting Information associated with this manuscript.

In addition to supporting students in class, the librarian also provided office hours to support student learning outcomes. At the start of each semester, the librarian hosted an “office open house,” where students were encouraged to stop by for informal introductions and conversations. Throughout the semester, students had several weekday opportunities to stop by for research support; these locations varied so that students could stop by an academic building that was convenient for them. From the Fall 2024 semester through Fall 2025, a total of 257 student interactions with the librarian were recorded. During these interactions, the librarian reinforced literature search skills, assisted with literature searches, helped to form search syntax, and instructed students on data management best practices.

An outgrowth of librarian involvement led us to recognize that students lack a differentiated understanding of web‐based information resources and cloud‐computing outputs. For many students, the process of discovering and citing information found through a web search tool like Google is familiar from other classes, but databases and domain‐specific tools are largely foreign. Once introduced to cloud‐computing resources, students express confusion about where information comes from, how it is stored, and how new data is generated [48, 49]. This general uncertainty leads students to question which resources are helpful for meeting their research needs and how to cite their sources correctly. It is important to reinforce to students that knowledge bases, such as UniProt, and library databases, such as SciFinder‐n, Scopus, and PubMed, provide web‐based information sources. The reference entries and journal articles found using these tools can be cited throughout the research project. As students progress through the projects and start to generate their own data, the librarian reinforces the concept that students are engaged in “real” research, whereby the use of computational software yields new results, which 1 day might be cited in a scholarly work. This issue arose especially when using cloud‐based computing platforms, such as Colab or AlphaFold3. Students have limited experience running computational software and, to students, these platforms provide an information output, but they do not see how this novel, unique‐to‐them data output differs from what can be obtained from a database. This is an opportunity to discuss the nuances of the research lifecycle in which they actively participate.

3.5. Expanded Use of Cloud‐Based Resources

As the project scaled to 100+ students, challenges with students using research‐grade software were exacerbated. Institutional computer labs at James Madison University are not set up for classes of more than 30 or for running scientific software needed for molecular dynamics. With the switch to remote and hybrid learning due to the COVID‐19 pandemic, these issues became more pronounced, accelerating the redesign of the project in this area. Table 1 shows the software and databases typically used in each project. Students receive instruction on using most of the software in class or via instructor‐created video walkthroughs.

For tasks such as model prediction and visualization, we used numerous free server options, which we incorporated into the project workflow (Table 1). For biomolecule visualization, we chose Mol*, a browser‐based program that works across a variety of web browsers and is accessible on mobile devices such as iPads, which many students use in lieu of a laptop computer [27]. However, there were activities that could be accomplished easily using ChimeraX or PyMOL that were difficult or impossible to perform in Mol* such as conversion of .cif files to .pdb files or making sequence changes [50, 51]. In addition, some webservers were unreliable or required a level of skill that was difficult to teach at scale to students. To address some of these issues, we turned to Google Colaboratory (Colab), a cloud‐based notebook that can run code written in languages like Python and R [20]. The Colabs run in a Chrome browser tab and do not use local resources outside the browser; thus, students can use them successfully on a variety of devices without coding knowledge. The Colabs can connect to a variety of GPUs, allowing for virtual access to high‐performance computing, but must remain stably connected to the internet. A number of scientific software or codes have been distributed as Colabs, including those for molecular dynamics with OpenMM [16, 28], protein structure prediction using Boltz1/2 and AlphaFold2 [17, 52], pK _a prediction [35], and ligand docking [53, 54]. We adapted or wrote the Colabs for molecular dynamics, ligand docking, pK _a prediction, docking visualization, normal mode analysis [29], and analysis of molecular dynamics [30, 55] to fill gaps in the broader project. While Colab generation requires some knowledge of Python or R, recent advances in LLM‐assisted coding and the incorporation of Google Gemini into the Colabs have lowered the coding knowledge needed to be successful for both instructors and students. Much of the work for the instructor was not developing the code but writing useful instructions for the students. For the class projects, we aim for Colab tasks to run in an hour or less so that they can be completed in class or when students are on campus and have more stable internet access.

3.6. Use of GitHub to Disseminate Project Data

For the final report, students are asked to generate a README file for their GitHub entry, which summarizes their data and results. This README is intended to be a summary of their final report, and a template is provided that indicates which sections of the report and data are needed. Previous students' README files on GitHub act as examples of what figures will be useful to collect throughout the semester, and which images of their modeled molecules can be best utilized to connect structure and function. One student issue that frequently arose in the final report and README file was students struggling to connect software to potential outcomes. In response, the instructor developed a homework problem set that asks students to connect data or results to the software used in class, along with a small introduction to writing the README file. The quiz helps students identify skills from earlier projects that may need reinforcement to succeed on Project 4. The instructor manages uploading the data to GitHub and only modifies the README files if the images or links do not render correctly. The instructor does not post projects that contain significant content or analysis deficiencies, such as missing report sections. This collection of projects is a public‐facing example of the students' work and can be included in student portfolios or on resumes/CVs.

3.7. Student Gains From Participating in the Project

Using a combination of public webservers and instructor‐generated resources, we have successfully implemented this project‐based curriculum in full since the Spring of 2024. In the current project iteration, 277 students were a part of the two course environments during the Fall 2024 and 2025 semesters and the Spring 2025 semester. In assessing students' confidence in their skills in analyzing molecular structures, connecting changes in DNA sequence to protein structure/function, forming hypotheses, making data figures, and drawing conclusions, we observe a shift from a large proportion of students needing constant guidance (light gray colors in Figure 1) to students showing independence (black color in Figure 1). The largest shifts in confidence occur when analyzing molecular structures and connecting them to the effects of mutations on protein structure and function, which aligns with the course project's structural focus. Further assessment of the skills in using literature offered similar increases in student confidence (Figure 2). Students showed large shifts in confidence in using databases such as Uniprot and in finding sequences, likely because this was most students' first exposure to these resources. In searching the literature, students showed continued development of these skills. Over half of the students felt they could confidently use Google Scholar and PubMed at the start of the project, and this number increased to nearly 75% by the end of the course. Google Scholar showed smaller confidence gains than PubMed, which may speak to the utility of this resource for broadly searching the academic literature and its introduction in other classes, while PubMed is more specialized to the biomedical literature.

FIGURE 1.

Student self‐assessment of skills—Students self‐assess their confidence in performing tasks associated with the project on a scale from 1: Could perform the task with step‐by‐step guidance from the instructor to 5: Could perform the task with almost no supervision or guidance. The complete scale with descriptions is listed in the methods. The skill for Qs. 1 was analyzing a molecular structure to determine chemical bonding and other properties, Qs. 2 making figures of structural or biomolecular data, Qs. 3 connecting a mutation in DNA to protein structure, Qs. 4 connecting a mutation in DNA to protein function, and Qs. 5 describing data and drawing conclusions based on data in a written format.

FIGURE 2.

Student self‐assessment of literature skills—Students self‐assess their confidence in performing tasks associated with using the literature and searching databases in the project on a scale from 1: Could perform the task with step‐by‐step guidance from the instructor to 5: Could perform the task with almost no supervision or guidance. The complete scale with descriptions is listed in the methods. The skill for Qs. 6 was searching for the biochemical basis for a disease in Uniprot, OMIM, or other curated databases. Qs. 7 finding the sequence of a protein in a database, Qs. 8 finding scientific manuscripts in PubMed, and Qs. 9 finding scientific manuscripts in Google Scholar.

4. Conclusions

Overall, incorporating a research project into a large biochemistry lecture course was successful, and we have addressed the challenges of integrating the project into the content typically found in a Biochemistry I course. Students feel more confident in their ability to find information, analyze data, and produce a public‐facing report of their work, which can be used as starting points for experimental verification. As the current project on modification sites in human malate dehydrogenases closes and we look to start the next topic, the framework described here will translate to projects with a variety of possible goals, including screening a library of potential inhibitors or comparing virtual docking and cofolding methods, such as those found in AlphaFold3 [24]. Algorithms like AlphaFold3 are also capable of predicting the structures of nucleic acids and complexes of biomolecules both with each other and with ligands. Thus, the framework outlined here could have students generate the model of a protein‐DNA complex to illustrate transcription factor binding to a promoter element followed by molecular dynamics of the biomolecules alone and the complex to detail the predicted effects of DNA binding on the transcription factor structure and dynamics. Students could then predict how DNA methylation or a variation in the transcription factor could alter the ability of the complex to form. In another example project, students could examine the effects of a small molecule on protein structure and function by predicting the complex by AlphaFold3 and a docking method, followed by comparing the structures in Mol* and comparing the simulated dynamics. Finally, the project could become more student‐driven, with students learning modeling approaches and then developing a novel project aligned with their interests, which may enhance student engagement and interest in the project and broader material. In all cases, the project presented here and the potential alternative directions can be linked to the content covered in introductory Biochemistry lecture and lab courses.

Funding

This work was supported in part by NSF MCB RUI‐2322867, DBI‐RCN 2119918, and the JMU Department of Chemistry and Biochemistry.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: Supporting Information.

Data Availability Statement

The data that support the findings of this study are openly available in the MCC MDH database at https://github.com/CEBerndsen/MCC_MDH_DB. Library lessons used in this project are available in the Supporting Information and at https://osf.io/yvm3q/overview.