Questions: RNA-seq Analysis Pipeline

Question 1

Why do RNA-seq aligners like STAR and HISAT2 need to be 'splice-aware'?

• A. Because RNA sequences contain introns that must be removed computationally
• B. Because mRNA has been spliced, so reads spanning exon-exon junctions will not align contiguously to the genome
• C. Because splice-aware aligners are faster than standard aligners
• D. Because RNA-seq reads contain adapter sequences at splice sites

Answer: (B)

mRNA molecules have had their introns removed by splicing, so the exon-exon junctions in the mRNA do not exist as contiguous sequences in the genomic DNA. When an RNA-seq read spans a splice junction, part of the read maps to one exon and part to the next exon, with an intron-sized gap in between on the genome. Standard DNA aligners would fail to map such reads. Splice-aware aligners like STAR recognize these split alignments and correctly place junction-spanning reads, which is critical for accurate quantification and isoform detection.

Question 2

TPM (transcripts per million) and FPKM (fragments per kilobase per million) always give identical gene expression rankings within a single sample.

• T. True
• F. False

Answer: True

Within a single sample, TPM and FPKM rank genes identically because they differ only by a sample-specific scaling factor. Both normalize for gene length and sequencing depth. The critical difference emerges in cross-sample comparisons: TPM values sum to one million in every sample, meaning that the proportion of expression attributed to each gene is directly comparable across samples. FPKM values do not sum to a constant across samples, making proportional comparisons between samples unreliable. For this reason, TPM is now generally preferred.

Question 3

Explain why raw read counts cannot be directly compared between genes of different lengths to assess relative expression levels.

Model answer: A longer gene captures more sequencing reads simply because it presents a larger target for random fragmentation and sequencing, not because it is more highly expressed. A 10-kb gene will accumulate roughly 10 times more reads than a 1-kb gene at the same expression level. Without normalizing for gene length, raw counts systematically overestimate the expression of long genes relative to short ones. Length normalization (dividing counts by gene length) corrects this bias, allowing fair comparison of expression levels across genes of different sizes.

This is a fundamental sampling bias in RNA-seq. During library preparation, RNA molecules are fragmented, and each fragment has an equal probability of being sequenced. Longer transcripts produce more fragments, hence more reads, for the same number of original RNA molecules. This is why all standard expression metrics (RPKM, FPKM, TPM) include a length normalization step.