Reading a genome browser

A genome browser is not a picture generator. It is a coordinate system: a reference sequence, features placed on that sequence, and evidence stacked as tracks. If you can say what each track is claiming, screenshots stop being decorative.

What must be in one window

Track	File you usually load	Question it answers
Reference	`FASTA`	What sequence are we numbering?
Genes / exons	`GFF` or `GTF`	Where is the gene model on this build?
Reads	`BAM` + index	Do the reads actually support this locus?
Variants	`VCF`	What alleles were called, and with what depth?
Signal	`BigWig` / bedGraph	Is coverage or ChIP signal continuous here?

Mixing builds (GRCh37 with GRCh38 annotations, for example) produces confident-looking nonsense. Write the assembly name in the same notebook as the accession.

A five-minute reading order

Read the contig name and the visible interval. Can you write chr7:55,190,000-55,220,000 without looking away?
Find the gene model. Are you in an exon, an intron, or intergenic sequence?
Look at read pileup. Soft-clipped ends, a sudden coverage drop, or a pile of low-MAPQ reads are arguments, not decorations.
Open the variant row at that position. Check REF, ALT, depth, and whether the allele is on both strands.
Only then take a screenshot — and caption it with the build, the interval, and the samples.

Common tools for the same exercise

The skill does not belong to one programme. Desktop browsers such as IGV and JBrowse are what most labs use. The UCSC Genome Browser is the usual web option for public assemblies. If you are checking a small local teaching set without a workstation, an on-device viewer that accepts the same FASTA / GFF / BAM / VCF / BigWig files — for example Japality Genomics Viewer — can show the identical coordinate problem. Keep unpublished files local; do not upload them to a random website just to “have a look”.

What students usually misread

Colourful coverage is not expression. DNA-seq coverage and RNA-seq coverage mean different things.
A gene arrow is a model. Alternative transcripts can place the “same” mutation in different exons.
Zero reads can be a mapping failure (repeats, decoy chromosomes, wrong reference) rather than a deletion.
A VCF row is a hypothesis. If the pileup does not support it, believe the pileup first and then debug the caller.

Write this in your notebook

Build: GRCh38 / hg38
Interval:
Gene / transcript:
What the BAM shows:
What the VCF claims:
What I still cannot decide:

Continue with Data formats, Alignment, and Variant calling.