📚 Syllabus & Course Curriculum
Bioinformatics & Computational Biology
Module-by-module breakdown of DNA Large Language Models (DNA-LLMs): Leveraging AI and NLP for Genomic Sequence Analysis, from foundations to a certified capstone project.
Adaptation
• Why DNA is not language: no words, no sentences, weak compositional grammar
• Tokenisation choices — single nucleotide, k-mer, byte-pair — and their consequences
• Context window against genomic distance, and the enhancer problem it creates
Models
• DNABERT, Nucleotide Transformer, HyenaDNA and Evo compared on context length
• Attention against state-space architectures for very long sequences
• Pretraining corpora and the species bias they carry
Prediction
• Promoter, enhancer and splice site identification
• Chromatin accessibility and expression prediction, and Enformer as the reference point
• Non-coding variant effect prediction where laboratory data is scarce
Practice
• Task heads, fine-tuning strategy and class imbalance in genomic labels
• Attention and attribution maps, and the weakness of reading biology from them
• In silico mutagenesis as a more defensible interpretation method
Evaluation
• Chromosome-level splits, since random splits leak through sequence homology
• Comparison against position weight matrices and CNN baselines
• Experimental validation such as MPRA before a regulatory claim is made
Earn government-registered certification in DNA Large Language Models (DNA-LLMs): Leveraging AI and NLP for Genomic Sequence Analysis
e-Certificate and e-Marksheet issued on successful completion.
