Space ships are always launched vertically. Its reason is pretty obvious: vertical launch has the lowest distance between the desired height and surface.
However, would horizantal launch be worth to experiment? Since space shuttles always landed on a runway, why we wouldn't launch them also on a runway, just like every other kind of airplanes?
Actually the main reason most launches begin vertically is to get away from the atmosphere, and atmospheric drag/heating as quickly as possible. This minimizes the amount of energy lost to friction, and the amount of mass needed to be used for heat shielding.
There is at least one orbital rocket whose first stage is an aircraft: the Pegasus. It's carried by a Lockheed L-1011. It's a small rocket, and, on a per kg basis, among the most expensive on the market. It survives, however, because it allows operators to put small satellites into non-standard orbits that wouldn't be reachable if they piggybacked on a large rocket also carrying a more conventional payload.
Virgin Galactic's SpaceShipTwo is also being piggybacked on an aircraft for its initial stage, but is only going for sub-orbital flights.
I suspect the main reason this approach isn't common is that a conventional, sub-sonic aircraft is able to impart only ~4% of the velocity needed to reach orbit. Which means you're adding a lot of additional complexity for a relatively modest saving in fuel/oxidizer consumption. Efficiency gains would be larger if higher speeds were achieved by the air breathing component; however, I believe the only manned ramjet engine aircraft to go beyond R&D prototypes into active service was the SR-71 Blackbird. While there also were about a dozen cruise missiles/SAMs/etc built using ramjets, as military projects, tech transfer to civilian industry is restricted.
It's possible that if the Sklyon, a winged SSTO with a hybrid ramjet/rocket engine is completed, the chicken and egg problem may be resolved; however as the Skylon is barely being funded at present, I suspect it will end up being nothing more than a paper study. If successful, the scramjet research being funded by DARPA and the AFRL might increase interest in launchers with air-breathing first stages by increasing their maximum velocity from mach 5 to mach 10; however in addition to limited funding there could be problems in transferring the knowledge from military to civilian custody.
Horizontal launch would be a great idea if you didn't have to deal with an atmosphere and had some altitude above the terrain as a starting point. This would allow you to burn perpendicular to the gravity vector for your entire ascent -- thus eliminating gravity drag completely. This is effectively what proposed mass drivers do for, say, lunar launches.
The question really has two questions, the first is well answered, about go straight up. Second question is about taking off from a runway.
The main benefit of a runway takeoff is to use aerodynamics to generate lift, requiring less thrust (thus less fuel) to get airborne and to a certain height.
The problem is that you run out of usable atmosphere for aerodynamic flight before you have gotten high enough to be useful. After that you need to become a rocket again. While you were in the atmosphere, you could capture oxygen to burn the fuel in, but once you get high enough you need to carry your own oxidizer.
The math just does not work out, to get a useful payload. The vehicle starts to be so big to carry all the fuel and oxidzer that the wheels, wings, other paraphernalia eat up any savings you might have gotten.
One idea that is a twist on this is to launch mostly empty, saving on takeoff mass, and refuel from a tanker, twice. Once for fuel, once for oxidizer then continue on.
