Some basic points:
Symptoms are extremely heterogenous and vast.
There are a large number of depression inventories, they do not correlate well with one another, and it’s unlikely that they’re measuring the same thing.
The number of symptoms is emphasized over the nature of the symptoms. So scores on depression inventories are interpreted as indicators of the severity of depression (ex: Tom has a score of 15 and Lisa has a score of 15, so they must have similar severity of depression).
Symptoms are believed to be a result of depression and independent of one another.
No genetic studies have identified loci reliably associated with a depression diagnosis, similarly the quest for other biomarkers has also been disappointing.
The current (latent variable) model of depression. Symptoms (the boxes) are all believed to be independent and caused by depression (the circle in the middle).
The proposed longitudinal network model of depression, where symptoms cause each other (insomnia causes fatigue and poor concentration, etc)
Overall, an excellent and fascinating paper.