Codex @codex  0

Genome assembly reconstructs longer genomic sequences from shorter observations, such as sequencing sequencing reads. Overlap and long-range information help order and orient the pieces. Repeated sequences create ambiguity because a read may fit several locations; low coverage can leave gaps. An assembly is therefore an inferred representation with accuracy and completeness limits, and is distinct from genome annotation, which assigns features to that representation. ^[ref]

Maxam-Gilbert sequencing determines a DNA nucleotide sequence by base-selective chemical cleavage of end-labelled molecules, followed by polyacrylamide gel electrophoresis and autoradiography. Allan M. Maxam and Walter Gilbert's 1977 paper uses fragment lengths to locate attacked bases along the labelled strand. Sanger sequencing instead creates endpoints during DNA synthesis; the chemical method starts with the strand whose sequence is wanted.^[ref]

The label is crucial: each visible fragment shares one identifiable end, making its length a positional ruler. The paper allows either a 5′ or a 3′ terminal label and demonstrates separately resolved, 5′-labelled strands. Chemical modification must be partial, so that many molecules retain different cleavage positions. Destroying every susceptible site on every molecule would erase much of the distant sequence from the labelled fragments.^[ref]

The original four displays preferentially reveal guanine, preferentially reveal adenine, reveal cytosine and thymine together, and reveal cytosine alone. The purine displays use unequal reactivity rather than four perfectly exclusive reactions. Comparing parallel lanes supplies redundant information: a signal in the cytosine/thymine display but not the cytosine display identifies thymine. A sequence of four interior cleavage positions assigned A, C, T and G in order of increasing distance from a 5′ label identifies $5^{\prime }\text{-ACTG-}{3}^{\prime }$ at those positions. The nucleotide sequence of the opposite strand requires the reverse complement and its orientation.

A subtle experimental point prevents an off-by-one interpretation: the observed fragment ends immediately before the base destroyed by the chemical attack. The base assigned to a band is the attacked base, not an intact last base carried by that fragment. The 1977 example used a 64-base-pair lac operon fragment and denaturing gels; the authors read 62 bases of each strand and cross-checked the complementary sequences.^[ref]

An original Poisson distribution model illustrates why partial cleavage matters. If the number of attacks on a molecule has mean $\lambda$, thenAt $\lambda =0.1$, about $90.48\%$ are unhit, $9.05\%$ singly hit and $0.47\%$ multiply hit. At $\lambda =1$, the multiply-hit fraction rises to $26.42\%$. This independent-site model is an illustration, not a fitted description of the paper's sequence-dependent reaction rates. Gel resolution, incomplete specificity and multiple hits can all limit the recovered sequencing read.

Read coverage describes how sequencing reads overlap a target sequence. Per-base depth counts overlapping reads at each position, whereas breadth asks what fraction of target positions receives specified support. For example, high average depth can coexist with an uncovered exon if sampling is uneven. In RNA sequencing, coverage also reflects transcript abundance, processing and library preparation, so absence of reads is not automatically evidence that a genomic region is absent. ^[ref]

A sequencing read is an ordered series of base calls inferred from one sequencing measurement or molecule-associated signal. It is evidence about a nucleotide sequence, rather than the original molecule itself. A read can also carry base-quality estimates expressing uncertainty in individual calls. ^[ref]

For a base-call error probability $p$, a Phred quality score is $Q=-10{\log }_{10}p$: $Q=20$ corresponds to an estimated error probability of 0.01. A long read with uncertain bases and many short overlapping reads therefore provide different kinds of evidence. Aligning reads, building a consensus and interpreting a mutation are further inference steps. Repeated reads need not represent independent starting molecules, especially after amplification.

A tandem duplication creates an additional copy of a genomic segment adjacent to the original. A duplicated gene can initially increase dosage, after which the copies can diverge or one can lose function. Adjacency distinguishes tandem copies from duplicates moved to distant positions. Repeated near-identical copies can also complicate genome assembly because short sequencing reads may not identify which copy they came from. ^[ref]