How you should change rooms in a game has a lot to do with the type of game you are using.  In one of the games I worked on all the walls where axis aligned rectangles so I could reuse each wall by calling functions which changed the location and size of the wall. Instead of having objects that are just sitting in memory not being used you can just burn a little cpu changing their values. 

draw(){

That's more or less what I would have suggested; however depending on the complexity of your game and the number of screens you have to deal with you may need more than just two main game states and so wouldn't be able to rely on a single boolean.  In that case I'd have a single state variable and set some constants to make the code more readable:

1. // constants
   
2. int PLAYING = 0;
3. int MAIN_MENU = 1;
4. int GAME_OVER = 2;
5. 
   
6. // globals
   
7. int gameState;
8. 
   
9. void setup() {
10.   size(600,400);
11.   gameInit();
    
12. }
13. 
    
14. void gameInit(){
15.   // Set the game's starting variables in a separate function which can then be called
16.   // again at game over to reset all necessary variables
    
17.   gameState = MAIN_MENU;
    
18. }
19. 
    
20. void draw() {
21.   switch(gameState){
22.     case(PLAYING):
23.       // main game loop
24.     break;
25.     case(MAIN_MENU):
26.       // display main menu
27.     break;
28.      case(GAME_OVER):
29.       // handle game over
30.     break;
31.   }
32. }

Alternatively if you do want to handle things in classes you could just use registerDraw() and unregisterDraw() to run the current game state's draw loop and stop the other states'...  In that case though you'd probably want to write a method to handle swapping game states.