Skip to content

Latest commit

 

History

History
109 lines (66 loc) · 3.07 KB

4-RPKM-FPKM-TPM.md

File metadata and controls

109 lines (66 loc) · 3.07 KB
Raw

R Notebook

RPKM FPKM and TPM

Replicates

Technical replicates

  • replicates from the same library with same techniques.

  • reduces random noise

Biological replicates

  • From biologically distinct samples

  • account for biological variations such as difference in temperature and environment…

Why do we have to normalize the count matrix from RNA data matrix?

It doesn’t account for a lot of biases.

Then which bias do you have to take account for

Biases

Gene Length

Assume Gene A is longer than Gene B.

Then more matches with happen in Gene A.

Sequencing Depth

Now we understand why we have to normalize

Normalization methods

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

    1. Normalizes the gene length and sequencing depth

    2. Higher than RPKM of a gene, higher the gene expression

    3. used to quantify transcripts from single-ended reads

    4. CAN NOT BE USED FOR DIFFERENTIAL GENE EXPRESSION ANALYSIS(deSeq2 / edgeR)

      • They take not normalized raw counts as a input

      • RPKM doesn’t account to some of the biases that happen in these techniques

  2. FPKM (fragments per kilobase of transcript peer million mapped fragments)

    1. Analogous to RPKM

    2. Higher the FPKM, higher the expression

    3. used for paired ended data, a read pair, rather than single reads

    4. FPKM ≠ 2 * RPKM

    5. CAN NOT BE USED FOR DIFFERENTIAL GENE EXPRESSION ANALYSIS(deSeq2 / edgeR)

  3. TPM (Transcripts per million fragments)

    1. Normalizes for gene length and sequencing depth

    2. TPM is better suited to compare expression between two samples

    3. CAN NOT BE USED FOR DIFFERENTIAL GENE EXPRESSION ANALYSIS(deSeq2 / edgeR)

Practice

6M library size for each replicates

Genes Gene Length Technical replicates 1 Technical replicates 2 Technical replicates 3
gene A 1.5 kb 50 25 85
gene B 2 kb 75 50 90
… … … … …
Total numbers mapped … 125 75 175

Calculate RPKM

  1. Normalize for sequencing Depth
    • Divide count / total numbers mapped RPM
    • for gene A = 50 / 125 = 0.4
    • However for real data you would have a really small number and that is why you multiply 1,000,000
  2. Normalize by Gene Length
    • normalize for 1kb of gene length RPK
    • 4 : 1.5kb = ?? : 1kb then 2.66 is the RPK value

TPM

  1. you do the RPKM but in reverse order

TPM is better to compare between technical replicates.