🔬 DNA Sequencing

Smith-Waterman is a dynamic programming algorithm for local sequence alignment, introduced by Temple Smith and Michael Waterman in 1981. It fills a scoring matrix where each cell score(i,j) = max(score(i-1,j-1) + substitution(i,j), score(i-1,j) - gap_penalty, score(i,j-1) - gap_penalty, 0). The zero floor is the key difference from global alignment: it allows an alignment to start anywhere and terminate at any high-scoring cell, making it ideal for matching short reads against a longer reference even when the read is noisy or only partially overlapping.

A SNP is a position in the genome where a single base differs between individuals or between a sequenced sample and a reference. In sequencing data, a SNP appears as a consistent mismatch across many independent reads at the same position. Variant callers distinguish true SNPs from sequencing errors by requiring the alternative base to appear in a minimum fraction of reads (typically 20–30%) and at sufficient coverage (often at least 8x). Random errors are scattered across reads and positions, whereas a true SNP recurs at the same site in every read that covers it.

N50 is the length of the shortest contig in the set of longest contigs that together cover at least 50% of the total assembly length. It is a length-weighted median rather than a simple average: a handful of long contigs can push N50 high even if many short contigs exist. In this simulation, higher coverage and lower error rates produce fewer, longer contigs and therefore a higher N50. The metric is the standard benchmark for comparing genome-assembly contiguity; a human genome assembly with N50 above 10 Mb is considered chromosome-scale.

Frederick Sanger and Walter Gilbert independently developed the first practical sequencing methods in 1977, sharing the Nobel Prize in Chemistry in 1980. Sanger's chain-termination method became the workhorse of the Human Genome Project, completed in 2003 at a cost of roughly three billion dollars. The launch of Illumina's short-read platforms in the mid-2000s reduced the cost per genome to under a thousand dollars by 2014. Oxford Nanopore and Pacific Biosciences subsequently introduced long-read sequencing, enabling near-complete assembly of complex genomes including the first gapless human reference sequence published in 2022.

Clinical genomics uses sequencing to diagnose rare genetic disorders, guide cancer treatment through tumour mutation profiling, and screen newborns for hundreds of inherited conditions. Infectious disease genomics enabled rapid SARS-CoV-2 variant tracking during the COVID-19 pandemic. Agriculture applies sequencing to breed disease-resistant and high-yield crop varieties. Forensic science uses short tandem repeat (STR) profiling for identification, and environmental DNA (eDNA) sequencing lets ecologists survey biodiversity from water or soil samples without capturing organisms.

Not necessarily. Coverage follows a law of diminishing returns: going from 5x to 30x dramatically improves assembly completeness and SNP accuracy, but going from 100x to 500x adds little for a standard diploid genome. Extremely high coverage can even introduce noise because sequencing artefacts — such as optical duplicates on Illumina platforms — accumulate and must be removed. For most germline variant studies 30x is standard; for somatic mutation detection in cancer, where only a fraction of cells carry the mutation, 100–1000x targeted coverage may be required.

Alignment maps reads to an existing reference sequence; assembly reconstructs a sequence from scratch without a reference using only the overlaps between reads. The overlap-layout-consensus (OLC) approach used in this simulation finds pairs of reads that share a suffix-prefix overlap, builds an overlap graph, finds a path through it (layout), and derives a consensus base at each position. De Bruijn graph assembly, used by tools such as SPAdes and Velvet, is an alternative that breaks reads into short k-mers and is better suited to short Illumina reads. Long-read assemblers such as Flye and Hifiasm can produce near-complete chromosomes in a single run.

Highly repetitive regions — centromeres, telomeres, and segmental duplications — remain difficult to assemble because reads shorter than the repeat unit cannot span them unambiguously; only ultra-long reads (100 kb+) from Oxford Nanopore resolve them. Haplotype phasing, determining which variants occur on the same chromosome copy, is still computationally demanding for heterozygous diploid or polyploid genomes. Epigenomic sequencing that simultaneously reads base sequence and DNA methylation from the same molecule is an emerging capability. Spatial transcriptomics, which maps gene expression to physical locations within tissue slices, and single-cell sequencing that sequences individual cells rather than bulk populations are rapidly expanding the scope of what sequencing can reveal.