I have some background in electronics and RF research, so I'll contribute what I can. The answer is Yes and No. (this is long, sorry)
For the "No:"
As far as the USB signal goes, the important thing is that the bits arrive properly and on time. Since there is some error correction in USB, one or two bits being lost isn't a big deal. Yes, there might be a few tens of microseconds of additional latency, but any USB interface will have a small buffer. So, as long as the cable is of reasonable quality, there's no problem.
Data problems then will be mostly limited to Jitter (since USB has error correction, and lost bits are very rare as it is anyway). A cable will NOT have an effect on the jitter. That's an issue with the clock. So get a reclocker (Hiface, etc, and give that clean power). A cable won't help you here.
Why? Given the physics behind the electrical signals, at 480 MT/s (megatansfers per second) on a single-channel serial link (USB 2.0), any reasonable cable <15 ft won't affect the timing. Make a ridiculous run of 100 metres unbuffered, and you might run into issues. At 10GT/s, yes, there are issues with reflection, timing, and such weirdness (hence with Thunderbolt's active cabling), and with parallel cables running at 100MT/s+. Neither of which apply to USB 2.0. Much less USB 1.1.
Now, for the "Yes:" Cheaper cables can introduce NON-DATA problems.
RF interference for one. A cable is an antenna. So an improperly shielded USB cable may introduce RF interference. Not anything at an audible frequency (the cable is much too short to get any apreciable low-frequency RF). However, higher frequency RF MAY introduce weirdness in the DAC module. So, moral is, make sure the cable is shielded (and the shield is grounded, which any decent cable will). Ferrites do help, in theory. I put my own on. I'm unsure if it made a difference, but they look cool. If you want to be extra cautious, get a USB 3.0 cable. There's more shielding on those.
Want to be extra sure? Make a homemade faraday cage and route as much of the cable through it. For extra peace of mind, just put the entire transport in a faraday cage, and make sure your building has a fancy matrix electrical earthing system, or that your hifi system has its very own grounding rod :)
But, the biggest problem is that USB comes with a 5V electric power feed. That is coming from a noisy place (your computer). This will affect most transports. Solution? Cut the red wire inside the USB cable, and feed in your own clean 5V power. You need the other three and the foil wrapping. Your cable has foil/metal braided wrapping, right? Any reasonable cable will.
Moral? Get a decent cable. No need for any ridiculous 30USD (internet price) things. What you get from any boutique ones (cough, Monster or Audioquest) is nice looks and bragging rights. The most important thing to remember here is that USB has error correction and any latency (microseconds) introduced will be nullified in the transport's buffer, if the buffer is properly compliant with standards.
But a consideration: placebo really is significant in the enjoyment of music. And with DBTs, there is a "confirmation bias" to watch out for. Science has that problem all the time. So from a psychologic and neurological standpoint, if it sounds "better" to use uber-fancy cables, go for it, even if it's simply due to placebo!
Edited by Chromako - 7/3/11 at 2:05am