RNA-seq Differential Expression, Alternative Splicing, Transcript Assembly and Gene Fusion

RNA-seq Differential Expression:

     Gene X’s expression in condition A doubles expression in condition B. But how reliable is this? What’s the chance of observing it by rendom? All comes to variation estimation! How to meassure the variance between different biological replicates. Once you have the variation estimation, you’re able to assign a p-value for expression changes. Variation can be estimated if you have many biological replicates.But in practice, we only have 2-3 replicates. What we can do next is proper statistical models.

Sequencing Read Distribution:

1. Poisson distribution:λ=E(X)=Var(X) 

The easiest model for RNA-seq reads count is Poisson distribution.

Assumption : Mean = Variance

But: sequencing data is over-dispersed,not only RNA-seq  (Mean<Variance)

2. Negative binomial: X ~ NB(r;p)( 2 parameters : r,p)

Definition : number of successes before r failures occur, if Pb(each success) is p .

Every gene has 2 parameters : mean and variance.

Negative binomial for RNA-seq : K_ij~NB(μ_ij,θ_ij²)

Variance estimated by borrowing information from all the genes - hierarchical models

Test whether μ_i  is the same for gene i between samples j.

Tools used in differential expression:

Cufflinks: versatile(RPKM/FPKM)

LIMMA-VOOM and DESeq: better variance estimates(TPM)

EdgeR; DESeq/DESeq2

Expression Index:

1. RPKM (Reads per kilobase of transcript per million reads of library)

    Corrects for coverage, gene length

    1 RPKM ~ 0.3 -1 transcript / cell

    Comparable between different genes within the same dataset

    TopHat / Cufflinks

2. FPKM (Fragments), PE libraries, RPKM/2

3. TPM (transcripts per million)

     Normalizes to transcript copies instead of reads

     Longer transcripts have more reads

     RSEM, HTSeq

Note: TPM is not able to compare expression of different genes across different samples. We should do differential expression on RPKM or FPKM

RNA-seq Alternative Splicing:

Different AS events:

 

TopHat: Assign reads to splice isoforms

MATS: Multivariate Analysis of Transcript Splicing

Transcript Assembly

Reference-based assembly: Cufflinks

1. Read mapping using Tophat

2. Construct a graph of reads

“Incompatible”fragments (reads) means they are definitely NOT from the same transcript

3. Identify the minimum # paths that cover all reads (each path is one possible transcript)

4. Transcript abundance estimation

De novo assembly: Trinity

De bruijn graph (1946):

Used in the earliest human genome assemblies

Standard algorithm for genome assembly

A sequence of length k can be represented as an edge between two sequences (length k-1)

Gene Fusion:

More seen in cancer samples

Still a bit hard to call

TopHatFusion in TopHat2