Although this is not a direct answer to your question, the comments would hopefully point you in the right direction. First of all I think you need to study the reflectance curves of minerals that you are interested in which are present in your study area. Once you know which minerals you are looking for, you can then overlay the mineral spectra (retrieved from the ASTER Spectral Library or your own laboratory measurements or In situ measurements) over the Relative Spectral Response Curve of the satellite sensor that you are interested in. USGS provides an example of a spectral characteristics viewer. Check the attached link. You can do something similar. Once you find a way to overlay the mineral spectra and the sensor response curves, you can look for the peaks and troughs in the mineral spectra which coincide with the sensor curves. Finally decide which band ratio would best work for your application.

http://landsat.usgs.gov/tools_spectralViewer.php

http://speclib.jpl.nasa.gov/search-1

 Hi, perhaps you can try Raster Calculator tools in GIS and 4 band divide 6 band (4/6 ratio). You will receive bitmap. Then you steel have to use Rasret Calculator and formula "bitmap>2" or "bitmap>1.5" (its up to threshold you want). Finally you will receive raster file which you can transform as a shape file, by using "Raster to polyline tools" (or Raster to polygon tolls. its up to you) and you will receive your serpentinites 

I don't hope so that can be done with multi-spectral data sets. Please go for hyper spectral data. You can download or pre-order EO-1 Hyperion data sets, that is available free of cost.