Project Title: Development of Web-Based Learning and Teaching Tools for Chemistry and Biochemistry

Project Lead's Name: Benjamin Gung

Project Lead's Email: gungbv@MiamiOH.edu

Project Lead's Phone: 513-529-2825

Project Lead's Division: CAS

Primary Department: Chemistry & Biochemistry

Other Team Members and their emails:

• Scott Hartley, hartlecs@MiamiOH.edu
• Jill Page, pagejs@MiamiOH.edu
• Mark Cybulski, cybulssm@MiamiOH.edu
• Alfredo Huerta, huertaaj@MiamiOH.edu

List Departments Benefiting or Affected by this proposal: Chemistry & Biochemistry

Estimated Number of Under-Graduate students affected per year (should be number who will actually use solution, not just who is it available to): 1,175

Estimated Number of Graduate students affected per year (should be number who will actually use solution, not just who is it available to): 10

Describe the problem you are attempting to solve and your approach for solving that problem: The goal of this project is to develop web-based tools to help teaching and learning fundamentals in chemistry and biochemistry in a more dynamic and interactive way. This goal will be achieved in part by developing a new interactive 3D-computer animation simulation focusing on atomic and molecular orbitals. In addition, we will develop new approaches to teaching and learning spectroscopy and organic reactions by making use of artificial intelligence (Al) and machine-learning tools. The overarching theme of this project is to stimulate student imaginations to improve the learning of difficult concepts in chemistry. This is a small budget, high payoff project for the students of Miami University.

In general chemistry (enrollment -800) and organic chemistry classes (enrollment -375), retention of students in the science majors remains a problem. Among other factors, many students struggle with some difficult concepts in chemistry and resist practicing homework problems which they do not consider interesting. In order to improve the retention rates and keep our students interested in important topics, we propose to develop new teaching tools in two areas.

1. The understanding of modern molecular orbital theory is fundamental in order to understand the structure and function of chemical compounds. The traditional way of introducing atomic and molecular orbitals has not been sufficiently effective and students often struggle with the concept, especially hybrid orbitals. Therefore how we teach our students in order to maintain their interest needs improvement. We intend to address this issue by developing a series of interactive 3D animations for teaching and learning. We will focus on topics such as the geometry and shapes of different atomic orbitals, the formation of hybrid orbitals, and the formation of different types of chemical bonds. We will use 3D modeling and animation tools (30S Max, ZBrush, etc.) to develop a suite of interactive animations that faculty members can use in the classroom and that students will be able to use for studying and learning on their own outside of the classroom. The interactive 3D animations will be made "playable" via a 3D gaming engine (Unity JD), allowing both professors and students to "play'' the animations and interact with them as needed for teaching/learning purposes. To make these animations more interesting and exciting to use, a "gaming" component will be included where appropriate, allowing students to learn each topic by "playing" for a higher score. Indeed, teaching and learning should be fun and these 3D animations will be a great tool to involve the students in a way not previously possible.
2. To learn organic spectroscopy (IR, MS, and NMR), students need to do a lot of problems. In spring 2013, we received a grant for the project entitled "Web-Based Organic Spectroscopy Problem Solving" from the Center for Teaching Excellence (then CELTUA). The goal of that project was to improve teaching organic spectroscopy to Miami students in several heavily enrolled classes. The organic chemistry classes that use this spectroscopy website include CHM254/255 (enrollment 75) and CHM244/245 (enrollment 300). That project still continues today by providing spectroscopy problems (http://chemistry.miamioh.edu/organicspecid/). That website has served our students and faculty members well in terms of convenience and availability of materials in teaching and learning spectroscopy. To encourage students to solve the spectroscopy problems on their own first, the answer key to those problems has been purposely kept confidential. We have traditionally received many emails from students every semester asking for the answer key to the problems. It is understandable that our students like to be able to compare their solutions to the correct answers, but for the benefit of future students, we still do not want the answer key to be exposed or published on the web. How can we satisfy current student interests, but also keep the answer key confidential? The answer to this problem is the second part of this project. We plan to take advantage of the rapidly evolving field of data science and artificial intelligence (Al). For any given topic, machine learning may be able to help in finding a pattern as long as there is sufficiently good data. We will develop a new Al-assisted website which will allow students to compare their spectroscopy solutions to answers provided by the Al algorithm. Students will no longer have to wait or ask the professor for the answer. They can now input their problems into the Al-backed website and get a suggestion. The Al-suggested solution does not provide all aspects of the correct answer but provides the key functional chemical group present in the unknown. This should keep the students interested in the problem-solving process and encourage them to develop critical thinking skills.

Another area in organic chemistry that takes practice to master is organic reactions. The new website we are building will include a reaction prediction site. When students input reactants and reagents, a machine-learning algorithm backed with a reaction dataset suggests correct reaction products. Students can compare their own predictions with the suggestion of the artificial intelligence program. Thus an otherwise laborious memorization process may become fun and interesting.

How would you describe the innovation and/or the significance of your project:

1. Animation of molecular orbitals and chemical bonding. The teaching of atomic and molecular orbitals is currently carried out in both general chemistry lecture courses, CHM 141/142 and the organic chemistry lecture courses, CHM 231, CHM 241/242, and CHM 251/252. The number of students enrolled in the general chemistry courses is approximately 800 in the fall semester (CHM 141), and 550 in the spring semester (CHM 142). The number of students enrolled in the organic chemistry courses is approximately 375 in the fall semester and 250 in the spring semester. Understanding molecular orbital theory concepts is a fundamental component of both general chemistry and organic chemistry. The design of truly effective educational visualizations of complex information is a multifaceted problem that must consider the multiplicity of learning styles presented by students (Scheiter, K. 2017, in Leaming from Dynamic Visualization Innovations in Research and Application. Springer 386 pp). To date, several learning models have been developed based on the use of complex Dynamic Visualizations. The Animation Processing Model (APM) proposed by Lo\1118 and Boucheix (Dynamic Diagrams: A Composition Alternative. In: Cox P., Plimmer B.1 Rodgers P. (eds) Diagrammatic Representation and Inference. Diagrams 2012. Lecture Notes in Computer Science, vol 7352. Springer, Berlin, Heidelberg) provides an effective principled and coherent approach for the development of educational animations. The five-phase APM model places special emphasis on the psychological effects of the visuospatial and dynamic character of animations, particularly on how they affect the learner's construction of a mental model A different model called the Composition Approach (Lo\1118, R. K., & Boucheix, J.-M. 2012. Dynamic diagrams: A composition alternative. In P. Cox, B. Plimmer, & P. Rogers (Eds.), Diagrammatic representation and inference (pp. 233-240). Berlin: Springer) relies on the design of animations that provide the student with a "simple-to-complex" series of components or "kits" that can then be used to construct a "whole" mental model. Other models of learning from graphical representations and dynamic animations have also been developed that focus on the many mental and psychological abilities of students. Our goal in this project is to make use of the current literature as a foundation for the design of 3D interactive animations that will be truly effective in helping students to develop clear mental models of the abstract but "real world" concepts that need to be taught in our courses. To demonstrate the general idea behind the use of interactive 3D animations, several of these have already been developed by Alfredo J. Huerta in preliminary form for demonstration, and two of them have been placed online for review (see below). Note that these interactive animations were prepared as "proof of concept" and will be modified and improved as we make progress on the project. The final version of the interactive animations will include additional models, explanatory text, voice narration where needed, and appropriate sound effects. As far as we know, nothing like these "interactive" 3D animations exists anywhere else for teaching these subjects. These animations represent a new teaching concept in that it makes use of a "gaming engine" for teaching selected topics in college-level chemistry and biochemistry.
   1. SPD atomic orbitals - https://huertaaj.itch.io/s-p-and-d-atomic­-orbitals
   2. Ethane Pi and Sigma hybrid orbitals - https:1/huertaaj.itch.io/ethene-molecule-double-bond-formation
2. Artificial intelligence assisted learning and teaching tools. The mastery of Nuclear Magnetic Resonance spectroscopy and familiarity with organic reactions is required for success in the organic laboratory and lecture courses. For that reason, it is important that these concepts are reinforced through practice problems, especially following a scenario similar to what students might expect in the real world. The task of keeping students interested in doing homework and practicing problems to achieve proficiency still relies on the assignment of homework and associated points. Student complaints continue to include words such as "too much homework", "not enough tutoring". We propose to use the progress made in modern data science and analytics, i.e. machine learning (for one of many books: Machine Learning for Dummies, IBM Edition, Judith Hurwitz, Daniel Kirsch, 2018, John Wiley & Sons Inc.) to provide help in student learning by collecting the relevant data (reaction data are in the textbooks) and find appropriate machine learning models. Once we have prepared our datasets and transformed the data into a computer-readable form, we can train the appropriate model algorithm to suggest the correct answers. Two different AI models will be used for organic spectroscopy and organic reactions, respectively. For the spectroscopy website, we are using the open-source machine learning library school-learn (see an entry in Wikipedia). It features various classification, regression and clustering algorithms and is designed to interoperate with the Python numerical and scientific library NumPy and SciPy. The functional groups in organic compounds are correlated to spectroscopic parameters such as wavenumber (cm-1) in IR and chemical shifts (ppm) in NMR spectra. A database for each type of spectra can be prepared by collecting real spectral data from various sources. Students can be trained to manually enter the data work with the functional group as the output (answers) and the spectroscopic data as the input (features). The resulting datasets are suitable for supervised learning in a classification model.

For the reaction prediction site, two recent publications in the Chemical Science journal forms the basis for our planned work. Both publications reported improved success rates applying machine learning to organic reaction predictions. One paper by Jensen et. al. used a graph of a convolutional network model for the prediction of reactions (Jensen et. al. Chemical Science, 2019, 10, 370.) Another paper by Schwaller et. al. used the NLP approach with better success in predicting complex reaction products (P Schwaller, et. al., Chemical science, 2018, 9 (28), 6091). We choose to adopt the NLP method because it is more straightforward in concept. To use an NLP (natural language processing) based algorithm, the open-neural machine translation library can be employed. (see Open-NMT-py, https://github.com/OpenNMT/OpenNMT-py). This is an open-source library for deep learning in natural language processing, originally developed for translations between two different languages.

However, most of the publically available reaction datasets were derived from the patent mining work of Lowe (Lowe, D. M. Extraction of chemical structures and reactions from the literature. Ph.D. thesis, University of Cambridge, 2012), where the reactions were described using a text-based representation called SMILES. Because the reaction dataset was derived from patent literature, the Schwaller model did not perform well with some simple textbook reactions such as SN1, SN2, E1, and E2 reactions. In this project, student assistants will be trained to manually enter textbook reactions for a new database. This involves using ChemDraw software to draw reactions and copy the drawings as SMILES formula. The SMILES formulae are letters and numbers which can be processed by the NLP program.

Using the deep learning neural network from Open NMT py with the new database, the resulting model should be able to generate correct answers to most textbook reactions. The training of the NLP models usually takes a long time (500,000 epochs, several days). Schwaller et. al. used an Nvidia Tesla P100 GPU and it took 48 hours to train their model using the US patent dataset. We tried the same dataset with an Nvidia 1080Ti GPU and it took 3 days to train the model. We have put this model on our website and the link is given below.

The AI-generated answers should elevate student interest and enhance student understanding of spectroscopy and reactions and should provide students with a modern learning experience.

Preliminary machine learning website under construction: http://organic.chm.miamioh.edu/machine_leaming/index.html

How will you assess the success of the project: Retention in related Chemistry courses will be used as one of the assessment tools. To assess retention in affected courses, we will compare the withdrawal rate from CHM 141/142 and persistence into the second-semester course with baseline data from the 2018-2019 academic year. Beginning in the second year, we will also track the number of students who have taken CHM 141/142 who remain in or switch to a major or thematic sequence in Chemistry or another STEM discipline.

The 3D animation of the atomic and molecular orbitals will be made available to students in CHM 141/142 and CHM 231/241/251 next year (AV 20/21). They will be utilized in classroom teachings with interactive animations in addition to traditional 2D images in their textbook/PowerPoint slides. Implementation of the project will make use of classroom time to introduce students to the 3D animation website and to demonstrate the use of the software. After this, the students should be able to access the online 3D animations on their own and use them for solidifying their understanding of the important concepts that are required for the course.

For the purpose of learning assessment, a quiz on this topic will be administered on Canvas after the introduction of the topic and the presentation of the 2D models. Then students will be introduced and allowed to play the simulations and take a second quiz on Canvas to assess whether or not they have gained mastery of the topic.

Since the 3D animations are being developed in an effort to make the orbital theory more approachable to the students taking the course, students will be polled at the end of each semester to determine their views on each of the individual animations. The results of the student polling will be utilized to help guide future changes in the course and improvement in the design of the interactive 3D animations where needed.

The new AI-assisted learning tools for CHM 242/252 will be tried out first by graduate TAs to assess the viability and timing of the tools. Starting in the spring of 2021, this project will be implemented in the latter weeks of the CHM 242/252 courses when students have familiarized themselves with common organic spectroscopy and reactions. The new machine-learning tools introduced during this project will include the use of retro-synthesis to determine what starting materials and reagents are needed to obtain the target compound. During this project, all of our available TAs will help introduce the new learning tools to undergraduate students in the courses mentioned above. Evaluation of this new tool will use a survey for students to fill out to determine the extent to which the project meets its overall goals. Based on feedback from the students the implementation will be revised to prepare for subsequent years in CHM 242/252.

Financial Information

Total Amount Requested: $10,596

Budget Details:

• Summer support for two students: $5,000
• Precision 5820 Tower Workstation with Nvidia Quadro RTX5000, 16GB GPU: (Dell website price $3,596) Miami Buyway says buy from a supplier. $3,596
• Software: $2,000
• Total budget: $10,596

Is this a multi-year request: No

Please address how, if at all, this project aligns with University, Divisional, Departmental or Center strategic goals: This project aligns with the University, Divisional, and Department strategical goal of improving our student retention rate. In general chemistry (enrollment -800) and organic chemistry classes (enrollment -375), retention of students in the science majors remains a problem. Among other factors, many students struggle with some difficult concepts in chemistry and resist practicing homework problems which they do not consider interesting. In order to improve the retention rates and keep our students interested in important topics, we propose to develop new teaching tools to stimulate student imagination and stay interested in chemistry and biochemistry.