Comp Bio Midterm

Contig

Continuous stretch of DNA sequence created by assembling overlapping sequence reads

Scaffold

Collection of ordered and oriented contigs, usually connected by “mate pairs”. Gaps between contigs in a scaffold are represented by unknown bases.

Coverage

Avg number of times each base in a genome was sequenced.

C = (# Reads * Read length)/Genome size

Contiguity goal

Maximize contig lengths and minimize total number of contigs

N50

Measure of assembly quality. 50% of the total assembled sequence is contained in contigs of this size or larger.

L50

Number of contigs whose lengths sum up to the N50 value

NG50

Similar to N50, calculated using the ESTIMATED total true genome size rather than the assembled assembly size.

Calculating N50

1) Sort contigs from largest to smallest

2) Sum all contig lengths to find the total assembly size

3) Divide the total size by 2 (50% threshold)

4) Keep a running sum of the sorted contig lengths. The length of the contig that makes the running sum meet or exceed the 50% threshold is the N50.

RNAseq Purpose

Used to reverse transcribe mRNA to cDNA and sequence it

RNAseq Uses

Used for measuring gene expression levels, differential expression, alternative splicing analysis, transcript discovery, annotation, and mutation discovery.

De novo assembly (alignment)

Used when there is no reference genome

Align to transcriptome

Best when you have known gene models

Align to reference genome

Use short tags to infer transcripts (requires mapping across splice junctions)

rRNA abundance (preparation challenge)

rRNA makes up ~90% of total RNA. Must either use poly-A selection to enrich for mRNA or rRNA depletion if studying bacteria.

RNA size variation (preparation challenge)

Small RNAs (<200bp) and long RNAs are sequenced separately

Stranded v. Unstranded (preparation challenge)

Stranded libraries allow you to differentiate between sense transcripts, antisense non-coding RNAs, and overlapping genes.

Same gene across samples (Quantifying Expression)

Important to control for sequencing depth. Not important to control for gene length.

Differentiate genes in the same sample (Quantifying Expression)

Must control for sequencing depth AND gene length (longer genes naturally produce more reads)

RPKM/FPKM (Quantifying Expression)

Normalizes for depth first, then length. Not perfectly comparable between samples because the mean expressed transcript length varies.

TPM (Quantifying Expression)

Normalizes for length first, then depth. The sum of TPMs in every sample is exactly 1,000,000, making it the preferred metric for comparing relative expression between samples.

Poisson Variance

Dominates at low reads (Technical noise)

Biological Variance

Dominates at higher expression levels

Replication

Always prioritize biological replicates over technical replicates (min of 3 recommended)

Depth v. Power

Deep sequencing helps find rare variants and low-abundance transcripts, but diminishing returns occur, and noise increases

Genome

~2% Protein coding genes

~20% Regulatory regions

Rest includes repeats and TEs

Ab Initio Prediction

Uses algorithms to predict genes strictly from the DNA sequence. Examples include Genescan, SNAP, and FGENESH. Algorithms look for ORFs, codon bias, start/stop codons, and splicing consensus sequences.

Evidence-Based Predictions

Use experimental data like sequenced mRNA or homologous protein sequences (BLAST) to find genes.

Automated Annotation Pipeline

Combine both ab initio and evidence-based models to produce a “consensus” model.

TopHat / STAR

Gapped mapper; map reads directly to a reference genome, allowing for gaps where introns exist.

Kallisto / Salmon

Extremely fast “pseudo-alignment” tools. Map reads directly to an existing transcriptome without doing base-by-base genome alignment, drastically speeding up transcript quantification.

Gene Ontology (GO)

Describes gene function across three domains:

1) Biological process (cell div and devel)

2) Molecular function (DNA binding and kinase activity)

3) Cellular component (nucleus, mem, etc)

General Databases

GenBank, RefSeq

Model Organism Databases

FlyBase, WormBase, SGD

ENCODE / modENCODE

Projects focused on cataloging functional regulatory elements (hypersensitive sites, ChIP-seq, WGBS, RNA-seq) across genomes.