Molecular Diagnostics AI Academy

Molecular Diagnostics AI Academy / AI in Biology

AI in Biology

How machine learning is used with sequences, images, text and laboratory data, how to tell whether a model really works, and where it can mislead. Written for biomedical, clinical and research scientists. No coding needed.

For biological scientists, biomedical scientists, researchers and undergraduate/postgraduate (UG/PG) students.

Modules
8
Study time
About 13 hours
Assessment
10 questions per module
Pass mark
8 of 10 (80%)
Cost
Free

Start the course

What you will cover

  1. Module 1: What AI can and cannot do in biology · about 95 min
    • Distinguish prediction from mechanistic explanation
    • Describe supervised, unsupervised and self-supervised learning with biological examples
    • Read AlphaFold confidence scores (pLDDT and PAE) correctly
    • Recognise the conditions under which machine learning tends to succeed or fail in biology
    • Frame a biological question as a machine-learning task with a stated intended use
  2. Module 2: Biological data and its traps · about 90 min
    • Describe the main biological data types and their typical artefacts
    • Recognise batch effects and confounding, and check for them before modelling
    • Recognise distribution shift between training data and the setting of use
    • Plan data quality checks and record provenance
  3. Module 3: Learning from data: classical models · about 90 min
    • Explain why training, validation and test data are kept separate
    • Explain overfitting, especially when features far outnumber samples
    • Describe regularised regression and tree ensembles and when each suits biological data
    • Use cross-validation correctly, with preprocessing and feature selection inside each fold
    • Compare a model against a simple baseline
  4. Module 4: Honest evaluation · about 100 min
    • Choose a data split that matches the intended use
    • Recognise the main types of data leakage
    • Choose and interpret metrics for rare outcomes
    • Explain calibration and why it matters for decisions
    • Report uncertainty and subgroup performance
  5. Module 5: Sequences and structures · about 100 min
    • Explain how DNA and protein sequences are represented for models
    • Explain what convolutional filters and attention learn from sequences
    • Describe how protein language models are trained and how their embeddings are used
    • Interpret structure-prediction confidence for domains, disorder and complexes
    • Split sequence data by similarity to avoid homology leakage
  6. Module 6: Microscopy and pathology images · about 90 min
    • Distinguish image classification, detection and segmentation
    • Score segmentation with IoU and Dice, and interpret the scores
    • Explain domain shift from stains, scanners and sites, and how to test for it
    • Prevent patient-level leakage in tile-based datasets
  7. Module 7: Text, language models and the clinical record · about 95 min
    • Explain how large language models produce fluent but false statements
    • Verify language-model output against primary sources, including citations
    • Identify special category data under UK GDPR and the conditions needed to use it
    • Choose appropriate uses of language models in laboratory and research work
  8. Module 8: Bias, reporting and regulation · about 105 min
    • Explain how bias enters through labels, proxies and missing groups
    • Audit subgroup performance
    • Use TRIPOD+AI to judge whether a prediction model study is completely reported
    • Describe the UK regulatory position for software and AI as a medical device
    • Write an intended purpose and a short model card

How it works

  • Who it is for: Biological scientists, biomedical scientists, researchers and undergraduate/postgraduate (UG/PG) students. Course-specific background: biomedical scientists, trainees, clinical scientists, researchers and students. You need basic statistics (proportions, sensitivity and specificity). No programming.
  • Each module: teaching, worked examples with real data, a case, interactive activities, common errors, a recap and referenced further reading.
  • Assessment: 10 questions, one at a time. Pass with 8 or more to open the next module. Retries use a different set of questions.
  • Study time is estimated from the finished lessons (reading at about 110 words a minute), plus activities, the assessment and further reading.
  • Certificate: available after all 8 module assessments are passed and payment is confirmed. Available for £99 after completion and payment. It shows your name, completion date and unique verification number, with CPD International accreditation.
  • Accreditation: Accredited by CPD International (cpdinternational.org).

Loading…