The lecture defines genomics as the study of genomes, organised around exactly three questions. Everything in the rest of the session sits under one of these three:

1. What is the sequence of the genome? — to read
2. What does the genome sequence mean? — to understand
3. How do we modify the genome sequence? — to edit

The lecture then puts up a single slide titled "Sharp decrease of sequencing cost" — and that slide is the whole economic story of this field. The $3 billion first genome is the reference point; everything since is a fall from it.

The lecture does not stop at listing platforms. It sets a discussion exercise: "Choose the most proper sequencing platform(s) — discuss with your neighbour classmate." And then it supplies the answer that matters in practice:

Whatever platform produced the data, the raw output has to be read. The lecture walks through the FASTQ format field by field — four fields per record:

The tool named for the first look at this file is FastQC — described on the slide simply as "a simple and useful quality check tool".

The lecture contrasts the two standard pipelines. They share a beginning and diverge at the middle:

De novo assembly is framed by two concepts the slides name: K-mer and De Bruijn graph-based assembly. The assembler table is worth memorising because read length is the selection criterion:

• Phylogenetics — comparing genomes across species. The lecture's worked example is the phylogeny of SARS-CoV-2, which is also a reminder that a phylogeny is built from sequence, not from phenotype.
• SNP and genomic diversity — single nucleotide polymorphisms as the substrate of human variation (cited to Science, 2008).
• GWAS — genome-wide association studies, comparing a case cohort against a control cohort. The slide points to the EBI GWAS catalog as the reference database.

The lecture makes a historical point that is easy to miss: annotating genes in model organisms started long before NGS existed. Researchers used genetic mapping to locate genes in organisms such as the fruit fly. Genes discovered in that era were mostly named after their function, usually by whoever found them — and some have funny names. The slide's editorial note: gene names are important, and if you discover a new gene you get to propose a name.

• Browsing genes — the lecture uses IGV (Integrative Genomics Viewer, igv.org) to view alignment and genome data, and GeneCards for gene records (the worked examples are PD-1 and PD-L1).
• Annotating a new genome — the method is similarity search: find the most similar gene sequence in an existing database. The tool is BLAST; for large-scale work you run BLAST locally, or use DIAMOND.
• Published annotations — NCBI datasets and the Ensembl FTP server (ftp.ensembl.org/pub/current_gtf/). Human annotation is distributed as GTF and GFF3 from Ensembl and GENCODE.

The lecture's light relief on this topic is a recent paper on "the lost tail of the human" (Nature, 2024) — the molecular answer to the monkey-tail question it opened with.

This is the commercial core of the deck: the choice between WGS, WES and a targeted panel is simultaneously a scientific decision, a cost decision and a product decision.

Variants are then classified by origin — germline, de novo and somatic — and the lecture uses the inheritance of sickle cell disease as the worked example of a germline single-gene disorder.

Types of genomic variation named on the slide: mutations, insertions, SNVs, deletions, tandem repeats and aneuploidy.

Detection methods are split into germline and somatic, and then into four methodological families: heuristic methods, statistical methods, deep learning methods and graph models.

Two further concepts are introduced here:

• Tumour mutation burden (TMB) — the number of mutations in a tumour, cited to Nature 2013 and NEJM 2017. This is a biomarker, and therefore a clinical decision point.
• The two canonical cancer genes — and the lecture's contrast is the cleanest way to remember loss-of-function vs gain-of-function:

  Gene	Class	Mutation type
  TP53	Tumour suppressing gene	Loss of function
  KRAS	Oncogene	Gain of function

• Risk allele frequencies and effect sizes — the lecture plots the relationship, which is the reason common variants usually have small effects and rare variants can have large ones.
• Case study: germline EGFR T790M and lung cancer risk (cited to Science, 2026) — a germline variant normally associated with resistance, appearing here as a susceptibility factor.
• Tandem repeat expansions — worked example HTT (Huntington's disease): <27 repeats = normal, 40+ repeats = disease (cited to Lancet, 2007). Note the shape of this: the same locus, the same mechanism, and the phenotype decided purely by repeat count.

Copy number variation (CNV) is where the lecture connects genomics to diagnostics. The key technical definition on the slide is the measurement itself:

Structural variations are treated as a separate class (cited to Chromosome Research, 2022).

The lecture gives translocation / gene fusion its own multi-slide block, and the reason is commercial: this is where a genomic finding becomes a drug.

• The illustration is Imatinib (Gleevec®) — a fusion-driven cancer treated by inhibiting the fusion product.
• Fusions are classified by tissue context: haematological, soft tissue and epithelial cancers — and within epithelial cancers the slide marks the range as rare to common (cited to Genome Medicine, 2015).
• For targeting, the lecture points to OncoKB as the actionable-gene reference.

The lecture closes the variant section with two slides headed "Our works", citing Cell 2018 and Nature Communications 2020 — the MET exon 14 skipping finding in secondary glioblastoma and the MGMT gene fusion driving temozolomide resistance respectively. Whether or not this appears in an exam, it tells you something useful: the lecturer works on exactly the kind of problem this lecture describes.

The final block answers the third genomics question — how do we modify the genome.

Then the lecture lands on the commercial endpoint: approved gene therapies, with prices. This table is worth knowing in full, because it is the clearest available picture of what a one-time genomic cure costs.

This session is unusually figure-heavy: 79 slides, and a large share of the substance sits in diagrams and tables rather than in body text. The block below recovers material that only becomes visible when the slides are viewed as pages rather than read as text — including one table that is genuinely load-bearing for anyone working in sequencing.

The lecture organises every major platform into three generations and compares them on seven axes. This is the reference table of the whole lecture.

The lecture then sets a worksheet: for each project below, state which platform(s) you would propose and why. This is the discussion exercise, and these are the five scenarios.

Variant calling is organised two ways at once: by the statistical approach, and by the variant origin. The lecture supplies the tool names for each cell.

This is the slide with the worked example, and it defines four mutation outcomes from one codon. Starting from the original sequence 5'-AAG-3' / 3'-TTC-5', transcribed to mRNA 5'-AAG-3', translated to Lys:

The slide draws the classification as a hierarchy, which is the cleanest way to remember it:

One slide earlier, the lecture pairs the central dogma (DNA → transcription → RNA → translation → protein) with the full genetic code table, with start and stop codons marked — the reference you need to read the table above.

The lecture's variation slide separates two levels, and it is worth learning the separation because exam questions tend to test one level at a time.

Rather than listing fusions, the lecture lays them on a century-long timeline and colour-codes them by tissue type — purple = haematological, green = soft tissue. The colour coding is the finding: the same mechanism recurs across unrelated tissues.

The lecture answers its own opening question ("How can we diagnose Down syndrome?") in the CNV section, and the answer is simply a copy-number change:

• Trisomy 21 — the karyotype shows three copies of chromosome 21 instead of two. That is the whole molecular definition.
• The slide pairs the karyotype with a clinical-features panel: small nose, mouth and ears; almond-shaped eyes with a skin fold covering the inner corner; white spots on the coloured part of the eyes; flat, round face; slightly bent fifth finger; short, stocky build.
• It also carries a caution worth keeping: intellectual disability and developmental delay vary by individual.

Method note: the tables above were recovered by rendering the slide pages as images and reading them, because a large share of this deck's content exists only as figures. Where a cell was legible on the slide it is reproduced as written; obvious typographical errors in the original (e.g. a manufacturer name misspelt, "expansive" for "expensive") have been corrected silently so the note is safe to revise from.

Why it matters commercially and clinically: the missing 8% is not junk — it is enriched for segmental duplications and repeat regions where structural variation lives, which is exactly where several disease-associated variants sit. Short-read resequencing could not see them.

Why this shows up in drug design: mRNA half-life is a design variable, and the 3′ UTR is one of the knobs. It also explains why mRNA level and protein level correlate poorly (the L6 number, r ≈ 0.4): two genes with identical transcript counts can produce very different protein amounts because their UTRs set different translation efficiencies and decay rates.

The experimental consequence: germline and de novo variants are found in blood or saliva; somatic tumour variants require tumour tissue (or ctDNA). Comparing tumour against the patient′s own germline is what separates a driver from a harmless inherited polymorphism.

None of these replaces the alignment and quantification tools covered earlier; they are the annotation layer you query after you have a result.

Why it is a good exam illustration: it is a non-coding change with a large phenotypic effect, it acts through splicing rather than through the protein sequence itself, and it shows an Alu mobile element doing real evolutionary work instead of being “junk DNA”.

It also mirrors a therapeutic idea already in the notes: if a disease is caused by exon skipping, then forcing or blocking that skip is a drug strategy — the same logic behind antisense oligonucleotides in splice-modulating therapies.