Lecture 5 - DNA Assembly

Lecture 5 - DNA Assembly

• The lecture covers DNA assembly, a bioinformatics process
• The goal is constructing a complete genome from fragmented sequencing reads
• The process involves several steps, starting with DNA sequencing, quality control, assembly, genome annotation, and analysis
• Various algorithms are used for DNA sequencing read assembly
• Coverage is an important factor influencing assembly accuracy
• Assemblies are often fragmented due to limitations in sequencing technology
• Repeats in the genome pose challenges for correct assembly
• Sequencing errors contribute to assembly fragmentation
• Diploid/polyploid organisms further complicate assembly

Lander-Waterman Equation

• An equation estimating the sequencing depth needed for genome assembly reaching a specific confidence level
• The equation (C = LN/G) considers genome coverage (C), read length (L), the number of reads generated (N), and genome size (G)
• Sequencing coverage is a measure of how many times a particular base is sequenced
• The calculation uses Poisson probability distribution

Lander-Waterman Limitations

• Assumes uniform coverage across the genome, but in reality, coverage varies
• Does not account for repeats, which are common in genomes. Repeats make it difficult to uniquely assemble genome regions
• Ignores sequencing errors that are introduced by the sequencing technology
• Assumes a haploid genome, despite most organisms having multiple copies
• Errors are not always random, the technology introduces biases

Genome Assembly Process

• The process involves assembling overlapping DNA sequences into contigs
• Contigs are then linked into scaffolds based on sequence similarities
• Scaffolds are positioned in pseudo-chromosomes based on physical location

Real-Life Complications

• Sequencing technologies introduce errors, these are context dependent
• Genomes evolve through duplication and deletion as well as mutation, making some regions difficult to uniquely identify
• It is not feasible to sequence every base in a large genome; not all parts are represented with the same frequency

Duplication Solutions

• Greater read length & read pairs provide improved results

Simple Repeats Solution

• Read length is the only practical solution at this stage
• Read pair distance is only a rough estimate, not an accurate solution

Genome Assembly Approaches

• Greedy assembly
• String graph assembly
• De Bruijn graph assembly

Greedy Algorithm

• A 'greedy' algorithm, Overlap-Layout-Consensus (OLC) is used to create assemblies
• OLC algorithms calculate nucleotide identity (overlap) between all pairs of reads
• A graph is constructed where each node represents a read and weighted edges between them represent their overlap quality

OLC continued

• Most OLC algorithms visit each node once
• Optimal path identification involves ordering and aligning overlapping regions of reads to construct a consensus sequence
• The complexity increases significantly with longer reads and more errors

Overlap Layout Consensus

• Example of how fragments share sequence in an assembly (layout graph of fragments)

String Graph

• Improvement over OLC algorithms (which is greedy)
• String graph algorithms examine and incorporate repetitive regions
• Algorithms use transitive reduction to compress repetitive regions

De Bruijn Graph Assembly

• Nodes represent kmers (short DNA sequences)
• Edges connect kmers that are adjacent in the original sequence
• Traversing the graph allows reconstructing the original sequence

De Bruijn Graphs and K-mer Size Impact

• Too small k-mer size leads to over-connectivity and collapsing of repeats
• Too large k-mer size leads to fewer overlaps and inaccurate assemblies

DBG Advantages

• Reduces computation time by not requiring explicit sequence-to-sequence overlap calculations
• The final sequence is generated by traversing the graph

Why are Assemblies Fragments?

• Sequencing technology limitations/weaknesses result in fragmented assemblies
• Complexity of the genome, particularly repeats, complicates assembly

Hamiltonian vs Eulerian

• Hamiltonian: Passes through each node once, but not necessarily all edges. More difficult to determine
• Eulerian: Passes through each edge once, but can visit a node multiple times. Algorithms exist to efficiently determine viability

Phased Assembly

• Separates maternally and paternally inherited chromosomes into haplotypes
• Helps with complex genome analysis of diploid or polyploid organisms
• Two main approaches exist: assembly-based phasing & alignment-based phasing

Genome Assembly

• Modern assemblies have four core steps (data pre-processing, contig construction, scaffolding, and gap closure)
• Specialized software handles these steps

The Perfect DNA Assembly?

• The current technology cannot achieve a perfect DNA assembly. Multiple conditions must be met (random errors, sampling is poisson, reads long enough to handle repetitions)

a) Errors Random

• As sequencing coverage increases, assembling errors become less impactful
• Non-random (reproducible) errors limit assembly quality despite increased coverage

b) Sampling is Poisson

• If sequencing coverage is sufficient, every region of the genome can be adequately covered.
• This relies on a Poisson sampling distribution

c) Reads Long Enough to Solve Repeats

• Sufficient read length helps resolve repeats and improves assembly
• Repeat resolution is crucial for accurate genome assembly

Can PacBio Achieve Perfection?

• PacBio technology has potential but challenges remain, particularly handling diploid/polyploid organisms, lengthy/noisy reads, and error rates
• Improvements in assembling error-prone reads are needed

Assembly with Long Reads

• Technologies focus on long read sequencing and assembly
• Three approaches include direct, hybrid, and hierarchical assemblies

Challenges of SMS de novo

• High error rates affect k-mer-based assemblies
• First whole-genome assemblies took significant computation time
• Newer software has improved assembly speed

DNA Minimizers

• Subsequences selected (kmers) to represent the read
• A smaller subset of kmers to reduce computational resources
• Minimizers are useful for identifying overlaps between sequences

DNA Minimizers 2

• Methods like BWT-FM kmer search to locate minimizers
• Minimizers for error-prone sequences

Minimizer DBG

• Minimizer DBG is a computational approach for assembling DNA sequences based on minimizers
• Utilizes error-corrected reads and minimizes the computational resources
• Can be used for assembly, error correction, and pan-genome assembly

Evaluating a DNA Assembly

• Metrics such as N50, NG50, L50 provide a measure of assembly quality
• Additional metrics include coverage and the number of contigs
• Tools for analysis include QUAST & BUSCO

Tools for Evaluating DNA Assembly

• QUAST: Quality assessment of assemblies. Provides metrics such as N50, L50, mis-assemblies, and completeness
• BUSCO: Benchmarking universal single-copy orthologs. Evaluates the quality of genome assemblies through assessing coverage by checking for the presence of highly conserved single-copy ortholog genes

This lecture focuses on DNA assembly, a crucial bioinformatics process for constructing complete genomes from fragmented sequencing reads. It explores various algorithms and factors affecting assembly accuracy, such as sequencing errors and genomic repeats. The Lander-Waterman equation is also discussed to estimate sequencing depth for assembly confidence.