Main scaling properties of the simulated galaxy samples compared to various observational datasets. Left panel: SFR vs M_⋆ for MHONGOOSE (dB24, blue squares), J0342-47 (light blue hexagon), HALOGAS (Heald et al. 2011; Kamphuis et al. 2022, cyan diamonds), and our TNG50 (orange circles) and FIRE-2 (dark red stars) samples. The blue dashed line shows the SDSS main sequence from Chang et al. (2015). The grey circles show the other systems from the MW/M31 TNG50 sample of Pillepich et al. (2024). Right panel: M_HI vs M_⋆. The markers are as in the left panel, with the addition of the SPARC sample (Lelli et al. 2016, green triangles). The purple-shaded area shows the ±1σ confidence region around the median scaling found by Maddox et al. (2015), while the blue dashed line shows the linear fit to the H I-WISE data from Parkash et al. (2018). Masses for the simulated galaxies are computed within 100 kpc from their centre. Black-edged markers are used for systems that are studied in this work. For comparison, we also show the simulated galaxies with H I masses computed with the recipe of Gnedin & Kravtsov (2011) (empty stars and circles).