Hi, filtered NOE removes resonances of protons covalently connected to other nuclei with non-zero spin - usually 1/2, probably almost always spin 1/2 nuclei, for example 13C or 15N.

The experiments work by converting 1H magnetization of protons attached to those nuclei to a state where 1H magnetization is not directly observable - due to scalar coupling interaction with that connected heteronucleus.

In a simple language - the 1H magnetization is transferred to the heteronucleus, it then may be dephased with a pulsed field gradient and made unobservable for the remaining time of the acquisition.

In two dimensional NOE experiments the filter element can be applied to directly detected dimension or indirect dimension, or both.

Imagine you have a mixture of a 13C and 15N labeled protein and a peptide, a piece of DNA or some other molecule (not isotope-enriched) that binds the protein.

It is then possible to use doubly filtered measurement to help assign the ligand spectrum and singly filtered - to find NOEs between the protein and the ligand (however there will still be signals of typu ligand-ligand in the same spectrum and they may actually be more abundant and make analysis still quite complicated).

Also, it is possible to record a filtered NOE with a third dimension - frequency of the heteronucleus. The 3D experiment will disambiguate the "ligand-ligand" signals from the "labeled species -- igand" but the sensitivity will be much worse than with the 2D experiment.