One way you can approximate this behavior is to build a transmission line structure similar in geometry in the stackup. I've done this before for a USB cable (check out the USB design kit for an exmaple).
This isn't going to be perfect, but since the cable is a twisted pair, the separation distance between the pair is fairly consistent across the whole cable length, so it's a good approximation. Also, the conductors are square instead of round because you're building them in the stackup essentially as traces, but for the speeds you're probably looking at running across 50m, this approximation is going to be a minimal impact as well.
If you can get an s-parameter model of the cable from the vendor, that's the best solution, but these tips should be able to get you simulating...
If the model that Steve suggested works in your situation, that is good. If not, you can make your own model in LineSim as outlined in this technote:
This gives the steps to model a single signal cable. For twisted pair, add another trace. Adjust the separation between the traces after step 4. i.e. adjust the trace width to get the specified single-ended impedance and then adjust the separation to get to the specified differential impedance. You may need to iterate these two steps a couple of times to match both impedance values. The differential impedance is more important in this case.
The best way to generate the CAT-5 simulation model is to perform lab measurements using an instrument that can output S-parameters. As example you can use SPARQ – Signal Integrity Network Analyzer from LeCroy, Agilent 86100C Infiniium multi-functional analysis tool from Agilent or any TDR from Tektronix that runs IConnect application. Alternatively a VNA will do the job for you, but it will be more complex to calibrate it. You can measure 1 m of cable and then scale the model to whatever length you need to.
The 10Base-T standard has a model based on a lumped component Bridge-T circuit. Values are given for 100 Meters of TP Cat-5.