it is obvious that there are no previous publications related to this compound have you obtained, therefore I recommend making a dose-response curve for the optimum concentration in where reactive oxygen species are inhibited. The concentrations should be in mg / kg in animals.

before making a dose-response curve you should determine the acute oral toxicity in rats of your new compound (OECD guidelines, number 425). Or at least know how toxic is your product in some cell lines.

Finding out acute toxicity according to the OECD guidelines would be one of the solutions

There are great difference between in vitro and in vivo results. In vivo testing you have to establish the safety and the efficacy of your testing material. One method as you been told is establish the acute toxicity by any method the you choose. Then you have to try  at least three dose levels in vivo , Starting by the least toxic dose for example 1/7 of the LD50 value the next dose suppose to larger dose e.g. 1/5 of the LD50, and the highest dose is 1/3 of the LD50

The following has to be taking in consideration:

1-Always run a control (placebo) group of animals in parallel experiment

2- Each experiment should be done with at least 5 animals

3-Record  daily  any sign and symptoms of your animals including : body weight and water intake..

$- You can use daily  multi drug administration.... 

 In order to establish efficacy You have to establish again efficacy doe response.  This has to be done on in vivo in animal model. remember that   Efficacy depends on safety of the drug . If your  if the margin of safety is wide you can try now a dose respond curve and pick the smallest  dose  withe minimum side effects. proper dose with no or minimum side effects  

In a nut shell; you can't.... you must estimate from prior experiences...Although the answers above are correct in essence (i.e. toxicology study), many world-wide university animal ethics committees will not allow the use of an acute tox study prior to decent justification of any activity at all in vivo! The dose you chose to give an animal (as mentioned will be in mg(ug?)/kg) will also greatly depend on the route you give, and of course the volume applicable to the area of admin , which will obviously change your molarity and therefore solubility. If you are to use the most simple and direct route (for instance), such as a subcutaneous route, then a highly potent drug in the nanomolar range may be estimated to be given at 1mg/kg or lower.. less potent drugs at higher, such as 10mg/kg. A good way to begin is look at known drugs targeting the same disease OR receptors/protein with a similar or lower potency, and begin by using the same dose/route as commonly published using that drug... perform them both (your drug and the known drug at the same dose and route). then you can compare efficacy between the two and go from there. Ideally you chose a conc. range (say 1mg/kg, 5mg/kg 10mg/kg) in a pilot study to determine a dose response, and determine a suitable 'maximal effective dose", based on volume and solubility of the compound as well. It's always a gamble to pick the correct dosage in an in vivo study BASED solely on in vitro data; My advice is to choose a known drug to comapre to; use the same dose as that, then chose a dose less than that (say 2-5fold less) and one more than that (say 2 fold more). This way you will have 1/ justification based on a reference compound, 2/comparison to a reference compound  3/ a dose response study around the 'effective' dose 4/ a pre-determined route.

Then as for the maximal effective dose; you picj the lowest dose that gives the BEST response without causing adverse effects. ie if your 10mg/kg dose not perform better than your 5mg/kg dose, then the 5mg/kg ids the max effectuive dose.

Hope this helps!?

BEst