Hm, instead of changing the transform.position of the object I would go and apply the pos.values to transform.localScale - see
You might have to play around with the values though.
As for audio cue:
You can put a collider around the object that is moved / scaled by the hand as well as the other object. If you don't want them to actually stop each other, it is possible to mark one of them (! never both) as trigger, which changes the methods a little. Also give them a tag that helps you identifying the correct collision before executing any actions.
Anyway, you need to attach a script to one of the objects. In that you will call either OnTriggerEnter or OnCollisionEnter
And to play Audio, you will need to work with the respective Audio components. if you have no previous experience, this should help out well:
:) Have fun coding! (and sorry again for being late - I meant it, poking me on Twitter is a lot faster because of a different notification system)