Valine and lactate have been recognized as important metabolic markers to diagnose brain abscess by means of MRS. However, in vivo unambiguous detection and quantification is hampered by macromolecular contamination. In this work, MEGA-PRESS difference editing of valine and lactate is proposed. The method is validated in vitro and applied for quantitative in vivo experiments in one healthy subject and two brain abscess patients. It is demonstrated that with this technique the overlapping lipid signal can be reduced by more than an order of magnitude and thus the robustness of valine and lactate detection in vivo can be enhanced. Quantification of the two abscess MEGA-PRESS spectra yielded valine/lactate concentration ratios of 0.10 and 0.27. These ratios agreed with the concentration ratios determined from concomitantly acquired short-TE PRESS data and were in line with literature values. The quantification accuracy of lactate (as measured with Cramér-Rao lower bounds in LCModel processing) was better for MEGA-PRESS than for short-TE PRESS in all acquired in vivo datasets. The Cramér-Rao lower bounds of valine were only better for MEGA-PRESS in one of the two abscess cases, while in the other case coediting of isoleucine confounded the quantification in the MEGA-PRESS analysis. MEGA-PRESS and short-TE PRESS should be combined for unambiguous quantification of amino acids in abscess measurements. Simultaneous valine/lactate MEGA-PRESS editing might benefit the distinction of brain abscesses from tumors, and further categorization of bacteria with reasonable sensitivity and specificity.