2.4.2. MEGA-PRESS data

Pre-processing and fitting was achieved using a combined FID-A [42] and Gannet [43] pipeline. Specifically, the data were concatenated and pre-processed using FID-A (run_megaoressproc_auto script) before being zero-padded and filtered (3 Hz line broadening) to match Gannet's pre-processing. Finally, it was fitted and quantified using Gannet (GannetFit, GannetCoRegister, GannetSegment, GannetQuantify). The following criteria were applied to excluding poor quality spectra: NAA linewidth >10 Hz, and a GABA:tCr ratio of <0 or >1. MRS data quality information for included participants is provided in Supplementary Table 1. This dataset was not included in the Glx analysis due to contamination/ signal distortion from poorly phased residual water signal.

Example Semi-LASER and MEGA-PRESS model fits are shown in Supplementary Figs. 18 and 19. Small differences in the TE and TR between the sequences will potentially lead to small differences in our quantification of GABA and Glutamate. However, these would be expected to be in the order of a few percent, in the worst case [44,45], and should be accounted for by modelling participant as a random effect in our model.