I am a chemistry graduate so may be able to help on the science side...but firstly..
This could be done. I wouldn't expect amazing accuracy nor precision. The resolution may be pretty good though (although useless).
You will need to do the following to form a spectrum:
A: A blank reading. A tube with the carrier gas an none of the target gas.
B: A calibration curve. A set of KNOWN concentrations of the gas you wish to measure.
C: A semi-wide spectrum light source of some form...ideally one that has a lot of its wavelength that is not absorbed by usual gasses in air (H20, N2,O2,CO2 mainly) but is absorbed strongly by your target gas. This will help with your sensitivity (signal:noise)...