According to ZWO, Gain of 350 should give you the least read noise for that model. Any gain setting above 200 is good - aim closer to 200 to get better dynamic range. Set exposure time as low as possible to get at least 50% (you can go lower than that but at expense of SNR), higher values are better (like already suggested 75% of histogram, provided you don't clip histogram or have too long exposure for given seeing). I usually leave gamma at 50 (it is only digital so no real benefit, and certainly so for planetary imaging), white balance at 50 for both blue and red (again no benefit since it is digital control and can be adjusted later after stacking). One thing to do as well - there is brightness setting (at least for 185mc, but I guess it is same for 224), when you decide on exposure length, cover telescope (as if taking dark frames) and look at the histogram. If it is clipping to the left, increase brightness setting. Minimum values of your dark frames should be just above 0. Do take at least 256 dark frames and process your recording in pipp (to do dark frame subtraction - this will remove both dark and bias signal).
For DSO there is no real benefit in going over 135 with Gain, so keep under. For shorter exposures stay above 60 for gain, for longer exposures you can go below 60 with gain. There seems to be a switch in read mode around gain value of 60 that has effect on read noise.
Last tip - don't use bias frames - take as many dark frames as you can at exactly same settings - without touching any of controls - just cover telescope with a cap.
It seems that these sony cmos sensors have some internal calibration that is applied whenever you change either gain or exposure length.