The DOI matches its Crossref record; 4 of the 4 references it cites were found in a registry or an index; 1 of the 3 sources its claims cite were checked for support.
Learning R with generative AI in a metagenomic data science course
Open accessAbstract
Generative artificial intelligence (AI) tools are increasingly used by students in introductory coding courses; however, evidence-based guidance for integrating these tools into biology education remains limited. We examined student experiences with generative AI in a beginner R programming course focused on metagenomic data analysis. An anonymous survey ( n = 43) captured quantitative ratings and qualitative reflections on how AI influenced learning, productivity, and problem-solving practices. Most respondents entered the course with little to no prior coding experience (79%) and reported frequent AI use throughout the semester, indicating that AI quickly became embedded in students’ workflows. Students rated AI as highly helpful for suggesting R code, explaining syntax and logic, and brainstorming analyses, with over 70% endorsing each use case. However, AI errors were common: over 90% of students encountered incorrect output at least sometimes, including domain-specific misinterpretations and overcomplicated or syntactically incorrect code. Notably, students identified a need for clearer instructional support in core AI-mediated practices. The most frequent recommendation for course redesign was to introduce foundational R concepts prior to AI use, highlighting a threshold-competency principle for effective AI integration. Together, these findings suggest that generative AI can support novice coders but does not substitute for foundational instruction. Effective AI integration requires deliberate pedagogical scaffolding and reflection rather than code generation alone. These principles are likely to remain critical as AI tools become more capable and more widely adopted in undergraduate biology education.
Overview
CoChat Summary
TL;DR
In a beginner R course focused on metagenomic data analysis, generative AI (Copilot) was widely used and rated helpful, but incorrect outputs were routine. Students recommended foundational R instruction before AI use and explicit support for prompting, debugging, and verifying AI output.
Key findings
- Most respondents (79%) had little to no prior coding experience; the majority used AI daily or several times per week, 89% found AI-use instructions clear, and over 70% rated AI as “very helpful” or “extremely helpful” for suggesting R code, explaining syntax/logic, and brainstorming NEON metagenomic analyses. Most estimated AI saved 2–3 hours per assignment, and overall effectiveness averaged 8.6/10.
- Over 90% encountered incorrect AI output at least “sometimes.” Common error types were misinterpretation of MAG and soil chemistry data (79%), overcomplicated solutions (70%), and incorrect R syntax or arguments (65%).
- When AI was wrong, students asked follow-up questions (95% of those encountering errors), ran/tested code in R (88%), consulted classmates (56%), checked documentation/help pages (49%), and consulted the instructor (44%). Students wanted more support in debugging with AI (65%), verifying AI output (63%), and prompting strategies (51%).
- Thematic coding of open responses found AI most often helpful for error diagnosis/debugging (39% of substantive responses), step-by-step explanations of code logic (18%), and code generation for novel analyses (16%); failure modes included repetitive incorrect suggestions (21%) and hallucinated variable or column names (17%). The most common redesign recommendation was to introduce R fundamentals before AI assistance (22% in the conclusions; Box 1 states 30%).
Methods
- Single-course anonymous survey (n = 43) in a beginner R programming course focused on metagenomic data analysis. The course consisted of 38 senior undergraduate Biology majors and six graduate students; Copilot was used because it was available through UMass accounts.
- AI-assisted coding was introduced in week 2. In the last 2 weeks, students worked with National Ecological Observatory Network metagenomes sequenced by the Joint Genome Institute, joined with soil chemistry and geolocation data, and were asked to “Vibe Code” by describing desired analyses for AI to generate.
- The survey used Google Forms and captured quantitative ratings plus qualitative reflections. Open-text responses were coded thematically across all four open-response questions; full response tables are in Appendix S1 and coding in Appendix S2.
Limitations
- Stated: findings are based on student self-reports, and most students had no prior experience learning R without AI, described as a key limitation. The paper also states it is not clear whether confidence gains are attributable to AI, and further research is needed to determine whether AI-assisted coding improves students’ ability to independently solve computational problems.
- Evident: this is a single-course, small convenience sample (n = 43) with no control or comparison group, so the survey cannot establish causality or whether efficiency gains reflect deeper learning versus effective delegation. Instructor observations about fewer coding questions and faster completion are also subjective.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
1 Valid · 1 Not valid · 1 Not checked · 38 Own results and interpretations
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Most students estimated that AI saved 2 to 3 hours per assignment (Fig. S5).”
§ Learning R with generative AI in a metagenomic data science course > CONCLUSIONS
hedged- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Importantly, most respondents indicated that they validated AI-generated outputs rather than relying on them blindly (Fig. S4).”
§ Learning R with generative AI in a metagenomic data science course > CONCLUSIONS
hedged- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Around 80% reported that AI improved their ability to learn from errors efficiently.”
§ Learning R with generative AI in a metagenomic data science course > CONCLUSIONS
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Most respondents entered the course with little to no prior coding experience (79%)”
- Also stated in: § CONCLUSIONS · plain
“Most respondents (79%) reported little to no prior coding experience, with many identifying as beginners (Fig. S1).”
§ Learning R with generative AI in a metagenomic data science course > Abstract
plainn = 43- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § CONCLUSIONS · plain
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“found the instructions on AI use to be clear (89%) (Fig. S3)”
§ Learning R with generative AI in a metagenomic data science course > CONCLUSIONS
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
8 more claims
Create a free account to see all 41 claims and every source verdict.
- Probably not a substantial claim. 12
- It may state two separate claims as one. 20
Citations
Cites
Shihab, M. I. H., Hundhausen, C., Tariq, A., Haque, S., Qiao, Y., & Mulanda, B. W. (2025) · Proceedings of the 2025 ACM Conference on International Computing Education Research V.1 · doi:10.1145/3702652.3744219 · Cited by 28
How AI agents will change research: a scientist’s guide
Gibney, E. (2025) · Nature · doi:10.1038/d41586-025-03246-7 · Cited by 12
AI is saving time and money in research — but at what cost?
Fieldhouse, R. (2025) · Nature · doi:10.1038/d41586-025-03936-2 · Cited by 5
Opportunities and considerations for using artificial intelligence in bioinformatics education
Piccolo, S. R., Nathan, A., Brazas, M. D., Kandpal, M., Miró-Herrans, A. T., Kleinschmit, A. J., McClatchy, S., Mutheiwana, P., Nikolic, D., Gallo, L. I., et al. (2024) · Bioinformatics Advances · doi:10.1093/bioadv/vbaf169 · Cited by 2
Metadata
- Authors
- Blanchard, J. L.
- Affiliations
- University of Massachusetts Amherst (US)
- Corresponding author
- Blanchard, Jeffrey L.
- Venue
- Journal of Microbiology & Biology Education
- Publisher
- American Society for Microbiology
- Published
- 2026-08-27
- Volume
- 27
- Issue
- 2
- Pages
- e0034325-e0034325
- ISSN
- 1935-7877 (ISSN-L)
- 1935-7885
- PMC
- PMC13520712
- OpenAlex
- W7162550083
- Semantic Scholar
- 4ba0f883d733040298bcf73c097d2236d7e77f89
- Cited by
- 0
- References
- 4
- FWCI
- 0.00
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- cc-by-nc-nd
- creativecommons.org/licenses/by-nc-nd/4.0/ (vor)
- journals.asm.org/non-commercial-tdm-license (tdm)
- Locations
- Journal of Microbiology and Biology Education (publishedVersion, cc-by-nc-nd)
- PubMed Central (publishedVersion, cc-by-nc-nd)
- PubMed (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
- Topics
- Physical Sciences › Computer Science › Computer Science Applications › Teaching and Learning Programming
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Genetics, Bioinformatics, and Biomedical Research
- Social Sciences › Education › Artificial Intelligence in Education
- Keywords
- generative AI, R programming language, metagenomics, biology education, AI-assisted code generation, novice programmers, instructional scaffolding, metagenomic analysis, student perceptions
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Learning R with generative AI in a metagenomic data science course. https://app.cochat.ai/papers/10.1128/jmbe.00343-25
- The paper
- Blanchard, J. L. (2026). Learning R with generative AI in a metagenomic data science course. Journal of Microbiology & Biology Education. https://doi.org/10.1128/jmbe.00343-25
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.