OMICS ANALYSIS II

RNA-SEQ

• RNA-Seq can reveal the quantity of RNAs in biological sample at a given moment and condition

​

APPLICATIONS OF RNA-SEQ

• Alternative gene spliced transcripts
• Gene fusion
• Non-coding RNA
• Quantification comparison

​

IDENTIFICATION AND QUANTIFICATION

• Identification: Knowing the sequence composition of the DNA/RNA reads for alignment.
• Quantification: Based on the alignment to calculate the amount of reads aligned at the genes.

ATAGACCGCGATA

Gene

Identification

Alignment

Quantification

COULD WE COMPARE THEM?

33-3=30

COULD WE COMPARE THEM?

All are higher

COULD WE COMPARE THEM?

S.1 has significant higher quantification

COULD WE COMPARE THEM?

Prot.1 has significant higher quantification

COULD WE COMPARE THEM?

Prot.1 at S.1 has significant higher quantification

NORMALIZATION

NORMALIZATION

NORMALIZATION

RPKM

• If the length of a gene is longer, more reads can map on it.

NORMALIZATION

RPKM

Divided by the sum of counts within the sample

RPKM

Divided by exon length

FPKM

• The idea of FPKM is the same as RPKM. FPKM is used for paired-end library and RPKM is for single-end library.
• FPKM will count the paired reads for only one time.

Gene

Only count as 1

TPM

Divided by the gene length

TPM

Divided by the sum of counts within the sample from 1 step

COMPARING RPKM AND TPM

CORRELATION COEFFICIENT

CORRELATION COEFFICIENT

• Pearson correlation coefficient
• Linear model
• Spearmann correlation coefficient
• Non-linear model
• Remove outliers
• Euclidean distance
• Distance based calculation
• Value differences can be considered

CORRELATION COEFFICIENT OR�EUCLIDEAN DISTANCE

Correlation coefficient is based on the trend. In this case, the correlation coefficient will be nearly 100%.

Euclidean distance is based on the distance between two lines. In this case, Euclidean distance is still not small.

The key is “the amount is matter or not”.

QUANTIFICATION COMPARISON

DIFFERENTIAL EXPRESSION ANALYSIS

• Differential expression analysis means taking the normalized read count data and performing statistical analysis to discover quantitative changes in expression levels between experimental groups.
• It is the most widely used analysis method for comparative transcriptomics.
• It is also often used for searching the biomarkers.

HER2 receptor

HER2 receptor

WHEN ARE TWO VALUES DIFFERENT

REPLICATES INCREASE RELIABILITY

• More replicates, more reliable mean estimates

​

• Define a fold change threshold ? Is it significant?

FOLD CHANGE NOT INFORMATIVE ENOUGH

• The 3 proteins have the same fold change

​

• Significance of protein B > signif. of protein A > signif. of protein C

T-TEST

• Consider both variance between groups and variance within groups
• Test-statistic:
• p-value: how likely is it to get a certain value of t by chance
• significant, if p-value <= significance level alpha, alpha = 0.01

P-VALUE

MULTIPLE HYPOTHESIS PROBLEM

• Multiplicity problem:
• 10,000 hypotheses are tested simultaneously
• Increased chance of false positives

​

• Suppose
• none of the 10,000 proteins is differentially expressed
• We expect 10,000 * 0.01 = 100 with p-value < 0.01

Individual p–values are no significant findings

BENJAMINI-HOCHBERG CORRECTION

• Compute p-values of all tests (m=10)
• Order them in increasing order

Test if:

​

34

Significance Level:

Adjusted Significance Level

PERMUTATION-BASED FDR

MEASURED

PERMUTED

• False discovery rate estimated by counting hits on permuted data

Careful with technical replicates !!!

VOLCANO PLOT

FDR cutoff at

Sign. Level 0.05

LogFC

-Log(q)

USING FOLD CHANGE

Difference

-log p

-log p

Difference

LogFC

-Log(q)

-Log(q)

LogFC

VOLCANO PLOT

The normal cutoff is

​

Q-value ≤ 0.05 which means

-log(q-value) ≥ ~1.3

​

Log₂(FC) ≥ 1 ~ 2 or Log₂(FC) ≤ -1 ~ -2

ENRICHMENT ANALYSIS

• Fisher’s Exact Test

• Gene Ontology
• Biological process (GOBP)
• Molecular function (GOMF)
• Cellular component (GOCC)

Contingency table:

In pathway?

Significant ?

Is a particular pathway overrepresented in the set

of differentially expressed proteins?

Difference

-log p

Fatty acid metabolism

Significant

Non Significant

• Pathways
• KEGG

ENRICHMENT ANALYSIS

• Fisher’s Exact Test

Contingency table:

In pathway?

Significant ?

Enrichment analysis on categorical data

LogFC

-Log(q)

ENRICHMENT ANALYSIS

MULTIPLE SAMPLE TEST

• Two samples test
• Treatment vs Control
• Multiple sample test
• Time serious
• Drug treatment comparison
• SNP comformation

1D ANNOTATION ENRICHMENT

All proteins

Do ribosomal proteins show preferentially higher abundances compared to all proteins?

Ribosomal proteins

1D ANNOTATION ENRICHMENT

Enrichment analysis on numerical data

All proteins

• Two-sample test
• Benjamini Hochberg FDR
• In addition: Rank-based score between -1 and 1

Ribosomal proteins

Score = 2(R1-R2)/n = 0.85

EXAMPLE OF PCA

PRINCIPLE COMPONENT ANALYSIS

PCA – Linear model

tSNE – Non-linear model

UMAP – non-linear model

CO-EXPRESSION ANALYSIS

DIMENSIONALITY REDUCTION METHOD

CLUSTERING HEATMAP

SPLICE EVENT

GENE FUSION

NON-CODING RNA

NON-CODING RNA

EQTL:EXPRESSION QUANTITATIVE TRAIT LOCI ANALYSIS

RIBOSOME PROFILING

• RNA-Seq based translomics
• Detection and quantification of the RNAs which are under translation

APPLICATION OF RIBO-SEQ

• Measuring protein synthesis
• Identifying novel protein – small proteins
• Working with ChIP-Seq to study cell regulation

DE NOVO SEQUENCING

• Mapping the reads without reference sequences
• Applications
• Novel species
• Paleo genome or proteome
• Lacking correct sequence information

5. Using close species to generate annotations.

PALEO-GENOMICS