In addition to John's recommendations above, be aware that carboxylates (acetate, citrate) may interfere in the ~1700 region, especially if your protein concentration is low (i.e. weaker signal from protein amide I & II). Just for the protein amide bands (~1500-1700), you can try to focus on a narrow frequency range and see if subtraction is doable assuming you can find reliable spectral features common to your protein solution and reference solvent. Other buffer systems (devoid of interfering bands in the region) can be tried.