Annotated Genomes are not static: they are changed when new evidence that improves the sequence is discovered. That's why we have different genome builds (ex: hg19, hg38).
Thus, the RefSeqs can change with them. Because it's difficult to change the genome build in the middle of a research project, the "old" information is largely still available at NCBI. Because some research relies on these RefSeq numbers, they are maintained independently of genome build.
However, sometimes, you need to know the base positions relative to the rest of the genes on the chromosome. This is when it's appropriate to use the Genome Annotation dependent RefSeqs.
RefSeq is composed of a non-redundant set of sequences. They are curated and corrected as new experimental evidence is found. You can see where the submission is in the process by looking at the RefSeq Status Code:
The Genomic subsection shows the position of the gene relative to the RefSeq and provides links to GenBank, FASTA, and Sequence viewer.
Genomic RefSeqs always start with NG_. The differences between the genome Independent and Dependent RefSeqs is the range. The gene prediction will still be largely the same, but the position on the reference sequence will be different.
Independent of Genome Build:
Dependent of Genome Build:
Most human genes are transcribed to produce multiple versions (isoforms) of a transcript. These in turn are translated into proteins that can have slight variations. Each one of these transcript isoforms gets a table in the mRNA and Protein(s) subsection.
A table with links to the original sequences that contributed to the RefSeq Annotations, which includes genomic, mRNA and protein sequences.