Long story short: from DOS, you can't get hybridization.

Long story long:

Let us first discuss: what is hybridization? The term is not as cleanly defined as you might think from the gloss-over in introduction lectures.

The original definition stems from the valence bond theory associated with Linus Pauling. In this model, orbitals undergo a "promotion" forming e.g. four sp3 orbitals for carbon. These would then form four sigma bonds with the neighbouring atoms. However, when you check for such sigma bond orbitals e.g. with photoelectron spectroscopy, you will find no such thing and canonical molecular orbital calculations will show an according 3+1 orbitals situation:

https://www2.chemistry.msu.edu/faculty/reusch/virttxtjml/Jmol-11/methane-mo.gif

You can sort of "rescue" the original hybridization concept by performing a unitary transformation on the HF or DFT wavefunction under certain constraints (e.g. the NBO method) and you will get orbitals that look e.g. sp3-ish. However, since these orbitals are no longer solutions to Hartree-Fock- or Kohn-Sham-equations so they no longer have the property "energy" that is testable by photoelectron spectroscopy.

While this is clear with state-of-the-art calculations and measurements, the condensed matter theory developed on the fly while this wasn't fully explored yet. Since the hybridization term already existed, they just transferred it and here the hybridization of something is the composition of the valence band. So while both things are called hybridization here, they are derived completely differently.

Back to the core of your question: the DOS is an integration over the band structure, so when you have only the DOS, the resolution of what the bands are doing and how they were composed is lost. You need the full band structure, in the best case a combination of ARPES and a quantum chemical calculation, to understand it.

Thank you very much Jürgen Weippert