---
title: "Appendix H: Teaching with the Life Sciences AI Handbook – The Life Sciences AI Handbook"
source: "https://lifesciencesaihandbook.com/teaching.html"
author: "Bryan Tegomoh, MD, MPH"
last_updated: "2026-10-30"
doi: "10.5281/zenodo.22073135"
license: CC BY 4.0
generated-by: handbook-markdown-export
---

# Appendix H: Teaching with the Life Sciences AI Handbook

Graduate programs in computational biology, bioinformatics, pharmacology, biomedical engineering, and translational science, along with training grants and research groups, can teach from the Life Sciences AI Handbook without asking permission or paying a fee. The materials below are built for an instructor with limited calendar time: short session tracks, a map to the ISCB competency framework, journal club questions for every chapter, case discussions, and assessment items with answer points. The full 12-week seminar, with a grading rubric and a learner AI-use policy, is the [Course Syllabus Template](./appendices/course-syllabus.html).

These materials support local teaching. They are not an accredited course and do not replace a program’s own curriculum, biosafety, or research-integrity review.

## Who the Materials Serve

The sessions assume graduate-level biology or a quantitative background, not both. Programming is useful for the practical chapters but not required for discussion.

| Setting | Suggested starting point | Typical slot |
|----|----|----|
| Graduate course in computational biology or bioinformatics | Eight-session track, or the 12-week syllabus | Course unit or elective |
| PhD program orientation or training-grant seminar | Four-session track | Seminar series |
| Laboratory or department journal club | Chapter questions in the journal club guide | Weekly or monthly |
| Drug discovery and translational science programs | Sessions on targets, molecules, translation, and evaluation | Didactic series |
| Industry onboarding for scientists new to AI | Four-session track | Training block |

## Session Tracks

Each session runs 60 to 90 minutes: learners read the assigned chapter sections beforehand, the instructor frames the question for ten minutes, the group works through one case or one primary study cited in the chapter, and the session closes with one or two assessment items. The competency column names where a session gives learners practice; the program decides how to assess it.

### Four-Session Track

| Session | Question for the session | Read before | Case for discussion | ISCB competency |
|----|----|----|----|----|
| 1 | When is a model output evidence, and when is it a hypothesis? | [AI for the Life Sciences](./foundations/ai-for-life-sciences.html); [Evaluation Principles for Life Sciences AI](./foundations/evaluation-principles.html) | [Docking Benchmark Failure](./appendices/case-studies.html#sec-case-docking-benchmark) | D3, F3 |
| 2 | What does a designed or predicted structure prove? | [Protein Structure Prediction](./molecular/protein-structure.html); [Protein Design and Engineering](./molecular/protein-design.html) | [Protein Binder Design](./appendices/case-studies.html#sec-case-protein-binder-design) | A3, F3 |
| 3 | Can a model predict how cells respond to something it has not seen? | [Single-Cell Foundation Models](./cells/single-cell-models.html); [Perturbation Prediction and Virtual Cells](./cells/virtual-cells.html) | [Single-Cell Perturbation Forecasting](./appendices/case-studies.html#sec-case-single-cell-perturbation) | B3, D3 |
| 4 | What has to be true before a discovery result reaches patients or the public? | [Translational Evidence and Failure Modes](./therapeutics/translation-failures.html); [Self-Driving Laboratories](./automation/self-driving-labs.html); [Information Hazards in Capability Research](./governance/information-hazards.html) | [AI-Discovered Drug in Phase IIa](./appendices/case-studies.html#sec-case-rentosertib-phase-2a) | C3, J3 |

### Eight-Session Track

| Session | Theme | Read before | Case for discussion |
|----|----|----|----|
| 1 | Foundations and data | [History of AI in the Life Sciences](./foundations/history.html); [Biological Data Infrastructure](./foundations/data-infrastructure.html); [Foundation Models for Biology](./foundations/foundation-models.html) | [Structure Prediction as a Research Input](./appendices/case-studies.html#sec-case-structure-prediction) |
| 2 | Evaluation and benchmarks | [Evaluation Principles for Life Sciences AI](./foundations/evaluation-principles.html); [Benchmarks for Bio AI](./governance/benchmarks.html); [Practical Evaluation Workflows and Worksheets](./practical/evaluation-workflows.html) | [Docking Benchmark Failure](./appendices/case-studies.html#sec-case-docking-benchmark) |
| 3 | Proteins and antibodies | [Protein Structure Prediction](./molecular/protein-structure.html); [Protein Design and Engineering](./molecular/protein-design.html); [Antibody and Biologic Design](./molecular/antibodies.html) | [Protein Binder Design](./appendices/case-studies.html#sec-case-protein-binder-design) |
| 4 | Genomes and variants | [Nucleic Acid and Genome Models](./molecular/genome-models.html); [Variant Effect Prediction](./molecular/variant-effects.html) | [Variant Interpretation Inside the ACMG Framework](./appendices/case-studies.html#sec-case-acmg-variant-interpretation) |
| 5 | Cells and tissues | [Single-Cell Foundation Models](./cells/single-cell-models.html); [Spatial Omics and Tissue Models](./cells/spatial-omics.html); [Perturbation Prediction and Virtual Cells](./cells/virtual-cells.html) | [Single-Cell Foundation Model in Practice](./appendices/case-studies.html#sec-case-single-cell-foundation-model) |
| 6 | Targets and molecules | [Target Identification and Prioritization](./therapeutics/targets.html); [Small Molecule Generation and ADMET](./therapeutics/small-molecules.html); [Chemical Biology and Target Engagement](./therapeutics/chemical-biology.html) | [AI-Discovered Drug in Phase IIa](./appendices/case-studies.html#sec-case-rentosertib-phase-2a) |
| 7 | Translation and clinical evidence | [Diagnostics and Biomarker Translation](./therapeutics/diagnostics-biomarkers.html); [Clinical Trial AI for Translational Research](./therapeutics/clinical-trials.html); [Translational Evidence and Failure Modes](./therapeutics/translation-failures.html) | Learner-selected example from the translation chapters |
| 8 | Automation, reproducibility, and responsible release | [Self-Driving Laboratories](./automation/self-driving-labs.html); [Agentic Science Workflows](./automation/agentic-workflows.html); [Reproducibility and Open Science](./governance/reproducibility.html); [Information Hazards in Capability Research](./governance/information-hazards.html) | [AlphaFold 3 Restricted Release and Open-Source Response](./appendices/case-studies.html#sec-case-alphafold3-access) |

### Twelve Sessions or a Semester

The [Course Syllabus Template](./appendices/course-syllabus.html) sets out a 12-week seminar with weekly readings, discussion questions, assignments, a grading rubric, a learner AI-use policy, and a 6-week intensive version. Programs that start with the four- or eight-session track can move to it without changing readings.

## Competency Map

The International Society for Computational Biology’s competency framework, version 3, sets out 13 competency statements, A3 to M3, each defined by knowledge, skills, and attitudes, and maps the minimum level of each competency, expressed with Bloom’s taxonomy, to bioinformatics career profiles ([Brooksbank et al., 2024](https://doi.org/10.1093/bioadv/vbae166)). The handbook covers the competencies concerned with judging and using AI in biology; it does not teach programming or computing infrastructure.

| ISCB competency | Chapters that address it |
|----|----|
| A3. Work at depth in at least one technical area aligned with the life sciences | The molecular, cellular, and organismal chapters matching the learner’s field |
| B3. Prepare life science data for computational analysis | [Biological Data Infrastructure](./foundations/data-infrastructure.html); [Microbiome and Multi-Omics AI](./cells/multiomics.html) |
| C3. Have a positive impact on scientific discovery through bioinformatics | [AI for the Life Sciences](./foundations/ai-for-life-sciences.html); [Translational Evidence and Failure Modes](./therapeutics/translation-failures.html) |
| D3. Use data science methods suitable for the size and complexity of the data | [Foundation Models for Biology](./foundations/foundation-models.html); [Evaluation Principles for Life Sciences AI](./foundations/evaluation-principles.html); [Single-Cell Foundation Models](./cells/single-cell-models.html) |
| E3. Manage own and others’ data according to community standards and principles | [Biological Data Infrastructure](./foundations/data-infrastructure.html); [Reproducibility and Open Science](./governance/reproducibility.html) |
| F3. Make appropriate use of bioinformatics tools and resources | [Toolkit for AI-Augmented Bio Research](./practical/toolkit.html); [Protein Structure Prediction](./molecular/protein-structure.html); [Model and Dataset Index](./appendices/model-dataset-index.html) |
| I3. Construct, manage and maintain bioinformatics computing infrastructure of varying complexity | [Workforce, Compute, and Institutional Readiness](./governance/workforce-compute.html) (planning only) |
| J3. Comply with professional, ethical, legal and social standards and codes of conduct relevant to computational biology | [Information Hazards in Capability Research](./governance/information-hazards.html); [Reproducibility and Open Science](./governance/reproducibility.html) |
| K3. Communicate meaningfully with a range of audiences | [Practical Evaluation Workflows and Worksheets](./practical/evaluation-workflows.html) (the evaluation decision record); journal club presentations |
| L3. Work effectively in teams to accomplish a common goal | [Workforce, Compute, and Institutional Readiness](./governance/workforce-compute.html) |
| M3. Engage in continuing professional development in bioinformatics | [Emerging Frontiers in AI for the Life Sciences](./future/emerging.html) |

Competencies G3 (designing user-centric tools) and H3 (scripting and programming) need practical coursework the handbook does not supply.

## Journal Club Guide

The chapter is the briefing; the paper is the evidence. Choose one primary study cited in the chapter, assign it with the chapter, and give the group 60 minutes in three parts: what the study claims (10 minutes), whether its design supports that claim (30 minutes), and what the next experiment or decision should be (20 minutes). The chapter questions further down are written for that last part and can be used with any study the chapter cites.

### Starter Papers

Five starting points the handbook discusses in detail make good first sessions. Each is read with its chapter, and the question gives the middle 30 minutes a focus.

| Study | Read with | Design question |
|----|----|----|
| [Jumper et al., 2021, *Nature*](https://doi.org/10.1038/s41586-021-03819-2): AlphaFold 2 | [Protein Structure Prediction](./molecular/protein-structure.html) | CASP14 accuracy was competitive with experiment for many single-chain targets. Which questions about function, dynamics, and binding does that benchmark leave open? |
| [Lin et al., 2023, *Science*](https://doi.org/10.1126/science.ade2574): ESM-2 and ESMFold | [Foundation Models for Biology](./foundations/foundation-models.html) | What did scaling a sequence-only language model buy, and how would you test whether its representations help your own prediction task? |
| [Theodoris et al., 2023, *Nature*](https://doi.org/10.1038/s41586-023-06139-9) (Geneformer), read with [Ahlmann-Eltze et al., 2025, *Nature Methods*](https://doi.org/10.1038/s41592-025-02772-6) | [Single-Cell Foundation Models](./cells/single-cell-models.html) | One paper reports transfer to network biology tasks; the other found that deep perturbation models did not beat simple baselines on common metrics. What baseline and metric would settle the question for your dataset? |
| [Watson et al., 2023, *Nature*](https://doi.org/10.1038/s41586-023-06415-8): RFdiffusion | [Protein Design and Engineering](./molecular/protein-design.html) | Which design claims rest on computational filters and which on experimental characterization? What success rate would you plan around in a wet-lab budget? |
| [Boiko et al., 2023, *Nature*](https://doi.org/10.1038/s41586-023-06792-0): Coscientist | [Self-Driving Laboratories](./automation/self-driving-labs.html) | The demonstration was in chemistry with a language model as planner. What would need to be shown before the same pattern runs unattended in a biology laboratory? |

### Questions by Chapter

| Chapter | Discussion questions |
|----|----|
| [AI for the Life Sciences](./foundations/ai-for-life-sciences.html) | Which output in the study was treated as a result before an experiment tested it? What benchmark, experiment, or deployment would move it from hypothesis to evidence? |
| [History of AI in the Life Sciences](./foundations/history.html) | Which older method does the study build on, and is it credited? What made this method important now: data, benchmarks, or a biological decision? |
| [Biological Data Infrastructure](./foundations/data-infrastructure.html) | What provenance, metadata, and negative results were available to the model? Which labels were measured and which were computed? |
| [Biomedical Knowledge Graphs and Literature AI](./foundations/knowledge-and-literature.html) | How were entities resolved and sources weighted? Which claims would a reviewer need to trace back to a primary paper? |
| [Foundation Models for Biology](./foundations/foundation-models.html) | What was the pretraining data, and how close is the evaluation task to it? Did the model beat a simple task-specific baseline? |
| [Evaluation Principles for Life Sciences AI](./foundations/evaluation-principles.html) | Did the data split match the biological question (sequence family, donor, scaffold, time)? Where could leakage have entered? |
| [Protein Structure Prediction](./molecular/protein-structure.html) | How were confidence scores used, and do they apply to the region that matters? What about the protein (complexes, ligands, dynamics, disorder) lies outside what the prediction shows? |
| [Protein Design and Engineering](./molecular/protein-design.html) | Which properties were measured experimentally (expression, binding, activity, stability, specificity), and what was the success rate? How many designs failed, and were they reported? |
| [Antibody and Biologic Design](./molecular/antibodies.html) | Which developability liabilities were tested? Would the result hold across epitopes and formats beyond the one studied? |
| [Nucleic Acid and Genome Models](./molecular/genome-models.html) | Which cell type, tissue, and measurement define the function claimed? What perturbation evidence supports the prediction? |
| [Variant Effect Prediction](./molecular/variant-effects.html) | How would the prediction enter ACMG/AMP evidence, and at what strength? For which ancestries was performance measured? |
| [Single-Cell Foundation Models](./cells/single-cell-models.html) | Did the embedding outperform simple baselines on held-out donors and protocols? How were batch effects handled? |
| [Spatial Omics and Tissue Models](./cells/spatial-omics.html) | How do spot size, segmentation, and capture efficiency limit the claim? Was the result replicated on another platform? |
| [Cell Painting and Image-Based Phenotyping](./cells/cell-painting.html) | Which assay choices (cell line, dose, time point) shape the profile? Was the mechanism confirmed with an orthogonal assay? |
| [Histopathology AI](./cells/histopathology.html) | Was performance tested across scanners, stains, and sites? Is the claim a research finding or a diagnostic intended use? |
| [Microscopy and Cryo-EM AI](./cells/microscopy-cryo-em.html) | What ground truth validated the processed images? Could the model have introduced structure that is not in the sample? |
| [Perturbation Prediction and Virtual Cells](./cells/virtual-cells.html) | Were genes, combinations, cell types, or laboratories held out, and which? How did the model compare with predicting no change? |
| [Microbiome and Multi-Omics AI](./cells/multiomics.html) | Were samples matched in time across modalities, and how was missing data handled? What design would support a causal claim? |
| [Systems Biology and Multiscale Modeling](./cells/systems-biology.html) | Which parameters are identifiable from the data? What experiment would most efficiently falsify the model? |
| [Neuroscience AI and Brain Foundation Models](./organisms/neuroscience-ai.html) | What recording modality, species, and task bound the claim? Does the decoding result say anything about cognition, or only about the task? |
| [Aging and Longevity Biology AI](./organisms/aging-longevity.html) | Does the biomarker predict outcomes, or only correlate with age? What evidence would link an intervention’s effect on the clock to healthspan? |
| [Plant, Crop, and Agricultural AI](./organisms/agriculture-ai.html) | How was genotype-by-environment interaction handled? Were predictions tested in field trials across seasons and sites? |
| [Environmental and Ecological AI](./organisms/environmental-ecology-ai.html) | Does the method measure detection, occurrence, or abundance? How was measurement uncertainty carried into the conclusions? |
| [Virtual Organisms and Digital Biology](./organisms/virtual-organisms.html) | Which subsystem was modeled, and was it validated against independent data? What would a claim of organism-scale simulation require? |
| [Target Identification and Prioritization](./therapeutics/targets.html) | Does the target hypothesis state disease context, direction of effect, and mechanism? What human genetic or experimental support exists? |
| [Small Molecule Generation and ADMET](./therapeutics/small-molecules.html) | Were generated compounds synthesized and tested, and how many? Could docking artifacts explain the ranking? |
| [Chemical Biology and Target Engagement](./therapeutics/chemical-biology.html) | What evidence showed target engagement in cells? Which orthogonal assay would confirm the mechanism? |
| [Drug Repurposing and Combination Therapy](./therapeutics/repurposing-combinations.html) | Is the exposure achievable at the dose required in the new indication? What evidence beyond signature matching supports efficacy? |
| [mRNA, RNA, and Vaccine Design](./therapeutics/vaccines.html) | Which step did AI inform (antigen, construct, delivery), and what immune data tested it? What remains to be shown in humans? |
| [Cell and Gene Therapy AI](./therapeutics/cell-gene-therapy.html) | How were off-target effects nominated and then measured? Which product-level safety evidence is still missing? |
| [Diagnostics and Biomarker Translation](./therapeutics/diagnostics-biomarkers.html) | What is the context of use, and was the model locked before validation? Which of analytical validity, clinical validity, and clinical utility does the study reach? |
| [Clinical Trial AI for Translational Research](./therapeutics/clinical-trials.html) | Did AI improve an operational step, or is it presented as replacing randomized evidence? How would a regulator view the endpoint? |
| [Real-World Evidence and Biomarker AI](./therapeutics/real-world-evidence.html) | What target trial does the study emulate, and how was confounding handled? Would the endpoint definition survive audit? |
| [Translational Evidence and Failure Modes](./therapeutics/translation-failures.html) | Which decision comes next for this result, and does the benchmark measure what that decision needs? Where in translation would you expect it to fail first? |
| [Self-Driving Laboratories](./automation/self-driving-labs.html) | What did the closed loop optimize, and was the measured endpoint the property that matters? Which protocol and calibration checks bounded its autonomy? |
| [Robotic Lab Automation and Cloud Labs](./automation/cloud-labs.html) | Did the protocol transfer between instruments or sites, and with what acceptance testing? Were failures logged and reported? |
| [Synthetic Biology Design Tools](./automation/synbio-tools.html) | Was the designed function confirmed in the host, and at what scale? How were containment and DNA screening addressed? |
| [AI for Biomanufacturing](./automation/biomanufacturing.html) | Which critical quality attributes did the model target? What validation would a GMP process require before use? |
| [Agentic Science Workflows](./automation/agentic-workflows.html) | Which tools could the agent call, and which steps did a person approve? How was provenance kept for each result? |
| [Toolkit for AI-Augmented Bio Research](./practical/toolkit.html) | Was the tool used inside the scope in which it was tested? What license, cost, and data-privacy terms apply? |
| [Practical Evaluation Workflows and Worksheets](./practical/evaluation-workflows.html) | Was the intended use written before the comparison was run? Which of adopt, bounded use, revise and retest, or reject does the evidence support? |
| [Benchmarks for Bio AI](./governance/benchmarks.html) | Is the benchmark blind and prospective, or retrospective and self-reported? What capability does a high score fail to show? |
| [Reproducibility and Open Science](./governance/reproducibility.html) | Could another group reproduce the computation from released code, data, and weights? Could it reproduce the experiment from the protocol? |
| [Information Hazards in Capability Research](./governance/information-hazards.html) | Did the paper withhold any method or artifact, and was that decision explained? Who reviewed the release decision? |
| [Workforce, Compute, and Institutional Readiness](./governance/workforce-compute.html) | Which roles in the study’s team made the result possible beyond model access? What would your institution need to repeat it? |
| [Emerging Frontiers in AI for the Life Sciences](./future/emerging.html) | What would have to be true for this frontier to become routine practice? Which evidence standard should apply to it? |

## Case Discussions

The [Case Studies](./appendices/case-studies.html) appendix holds short cases across the book. A role-based format works well: assign learners the roles of principal investigator, computational lead, wet-lab lead, biosafety officer, and journal referee, and have each argue from that role before the group reaches a decision.

| Case | Prompts |
|----|----|
| [Structure Prediction as a Research Input](./appendices/case-studies.html#sec-case-structure-prediction) | What did the predicted structure make possible, and what did it leave to experiment? How should confidence be reported to collaborators? |
| [Docking Benchmark Failure](./appendices/case-studies.html#sec-case-docking-benchmark) | What made the benchmark look better than the method was? How would you design a test that the same artifact could not pass? |
| [Single-Cell Perturbation Forecasting](./appendices/case-studies.html#sec-case-single-cell-perturbation) | Which baseline did the model need to beat, and did it? What held-out condition would convince a skeptic? |
| [A-Lab and the Novelty-Claim Discipline](./appendices/case-studies.html#sec-case-a-lab-novelty) | What evidence should support a claim that something is new? Who should check it before publication? |
| [AI-Discovered Drug in Phase IIa](./appendices/case-studies.html#sec-case-rentosertib-phase-2a) | Which part of the program did AI contribute, and which parts were conventional drug development? What would a Phase IIa result establish, and what would it not? |
| [Agentic Research Planning Without Lab Authorisation](./appendices/case-studies.html#sec-case-agentic-research-planning) | Where should the line sit between an agent planning research and a laboratory executing it? Who authorizes the crossing? |

## Assessment Questions

The items below suit short-answer quizzes, oral examination prompts, or the closing minutes of a session. Answer points are collapsed so the page can be projected.

1.  Why should a model output be treated as a hypothesis until tested, and what can test it?
2.  Give two examples of data splits that match a biological question, and explain what a random split can hide.
3.  A structure prediction has high overall confidence. What can still be wrong for the question you care about?
4.  A protein design paper reports many plausible designs. Which measurements decide whether the designs are results?
5.  How should a computational variant-effect prediction be used in clinical variant interpretation?
6.  A single-cell foundation model reports strong transfer performance. What comparison should accompany that claim?
7.  Why is docking performance alone a weak basis for advancing a compound series?
8.  What distinguishes analytical validity, clinical validity, and clinical utility for a biomarker?
9.  Why are computational and experimental reproducibility separate requirements?
10. What should a research team weigh when deciding how much detail to publish about an AI-biology capability?

Answer Points

1.  A model can fit benchmarks or retrospective data without capturing the biology the decision depends on; a benchmark designed for the question, an experiment, or a deployment context tests it ([AI for the Life Sciences](./foundations/ai-for-life-sciences.html)).
2.  Sequence-family splits for proteins, donor splits for cells, scaffold splits for molecules, and time splits for prospective use. A random split can let near-duplicates into both training and test sets, inflating performance ([Evaluation Principles for Life Sciences AI](./foundations/evaluation-principles.html)).
3.  Local confidence in the region of interest, conformational state, disorder, and interactions with ligands, nucleic acids, or partners can all differ from the overall score ([Protein Structure Prediction](./molecular/protein-structure.html)).
4.  Expression, folding, binding, activity, stability, and specificity measured experimentally, with the success rate and failures reported ([Protein Design and Engineering](./molecular/protein-design.html)).
5.  As one line of evidence within ACMG/AMP interpretation, alongside segregation, functional, and population evidence, with attention to the ancestries in which it was validated ([Variant Effect Prediction](./molecular/variant-effects.html)).
6.  A comparison with simple baselines on held-out donors, tissues, protocols, and batches ([Single-Cell Foundation Models](./cells/single-cell-models.html)).
7.  Docking poses and scores can reflect artifacts; the series needs route feasibility, measured binding, selectivity, ADMET, and experimental validation ([Small Molecule Generation and ADMET](./therapeutics/small-molecules.html)).
8.  Analytical validity is whether the test measures the analyte accurately; clinical validity is whether the result tracks the clinical state; clinical utility is whether using it improves decisions or outcomes in the stated context of use ([Diagnostics and Biomarker Translation](./therapeutics/diagnostics-biomarkers.html)).
9.  Rerunning code on released data and weights can succeed while the wet-lab protocol, reagents, or instruments cannot be reproduced, and the reverse can also happen ([Reproducibility and Open Science](./governance/reproducibility.html)).
10. Whether specific methods, details, or artifacts meaningfully change access to harmful capability, balanced against the value of the evidence to legitimate research, with review before release ([Information Hazards in Capability Research](./governance/information-hazards.html)).

## Citing the Version Used in a Course

The handbook is revised continuously, so a syllabus should record the date on which learners read it. The [Changelog](./changelog.html) lists what changed after that date.

> Tegomoh B. *The Life Sciences AI Handbook: Evidence for AI in Biology and Drug Discovery*. 2026. Accessed \[date\]. doi:10.5281/zenodo.22073136

Chapter-level citations and other formats are on the [How to Cite](./how-to-cite.html) page.

## Reuse and Adaptation

The handbook text is licensed under the Creative Commons Attribution 4.0 International License ([Creative Commons, CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)). Programs may copy, excerpt, translate, and adapt it into slides, course packs, learning management systems, and assessments, including for courses that charge tuition, provided they give credit, link to the license, and note any changes. An attribution line can read:

> Adapted from *The Life Sciences AI Handbook* by Bryan Tegomoh, MD, MPH (lifesciencesaihandbook.com), licensed under CC BY 4.0. Changes: \[describe\].

Studies, figures, datasets, model weights, and quotations from other sources keep their own licenses; cite them from the original.

## Contact

Using this handbook in a course or program? [Tell us](mailto:bryan.tegomoh@gmail.com?subject=Teaching%20with%20the%20Life%20Sciences%20AI%20Handbook). A line on the program, the sessions used, and what did not work helps decide what to revise next.