I did try western blot but I have no luck with it. 

Antibodies don't always work well for Western blots if their affinities are too low. Another potential problem is that small proteins like yours may transfer right through the membrane if you're not careful. Shortening the transfer time and changing the type of membrane may help. Make sure the gel you are using separates small proteins like yours.

You could try sandwich ELISA if you have 2 different antibodies recognizing the protein with different epitopes.

You could bind your antibody to a surface plasmon resonance chip and flow your extract over it. 

You could also try a fluorescence polarization immunoassay (FPIA), which would require a fluorescently labeled peptide consisting of the epitope of the antibody (assumes a mAb). The protein of interest competes with the fluorescent peptide for binding to the antibody resulting in a decrease in fluorescence polarization.

In addition to the above suggestions, you could also try MALDI TOF-TOF (MS-MS) to detect your protein. when you run the results obtained in MALDI against a standard database (Mascot) your protein will be detected (provided it is there in the database) 

It is also possible that the protein of your interest can be identified by a targeted mass spectrometry-based approach known as multiple reaction monitoring (MRM). In this technique, there will be no requirement of using antibodies.

What kind of gel (and %) are you using for 8 kDa protein separation. Furthermore, are you optimizing this procedure for the first time or you lost what you have seen in the past. Generally, low molecular weight protein are difficult to run and transfer on membrane. Please elaborate more on your question (s) for a specific answer. Did you stain the membrane to see if the transfer conditions are right for you? I am assuming that you are getting low bands (markers) on the membrane (what kind?).

Lewis- different antibodies will work for different assays- many antibodies are known to work for western but for immuno-fluroescence or vice versa, so you might need to try your westerns again with different antibodies. The problem there is often that different sample treatments lead to loss of the epitope the antibody would detect, and westerns can be optimised for epitope preservation or retrieval if you are using a buffer transfer step b using different buffers, and by cooling buffers and running the transfer in the cold room for example- unfortunately probably no good if using the iBlot or other dry/semi-dry transfers.

If abundance of your target, or affinity of your antibody, are a problem, then you could try combining a pull-down using your antibodies immobilised on (e.g.) protein A beads with the western blot you are trying to do. By boiling the beads in sample buffer you run a sample that is both enriched for the target and reduced in terms of background, both of which will help. If your target is abundant enough you may even be able to see it on a Coomassie or silver stained gel without doing a western, and then cut the band out for unequivocal identification by mass spec. 

Finally- because there are so often problems with antibodies, you could also turn to a company like ours to make a non-antibody affinity reagent...

Paul

If you want to use western blot, you can work out conditions for the transfer using a pre-stained molecular weight marker that has a protein of the size in the vicinity of 8 kD. You may want to vary the percentage of acrylamide in the gel (higher than you have been using), time for transfer (shorter than the time you have been using), voltage (lower than what you have been using) or some combination of those changes.  When using pre-stained markers, it is easy to see how much has transferred and sometimes you may even see how much was left behind on the gel.  You can also increase the ability of your antibodies to detect the protein by using a biotinylated secondary antibody followed by streptavidin conjugated to the enzyme like HRP. 

Well, I am aware that some antibodies just don't work, and often you cannot trust the manufacturers, BUT before giving up your antibody I suggest that you try some modifications in your protocol.

for small proteins/peptides I would use a 16.5 % acrylamide (or a 4-20 % acrylamide) with Tris-Tricine.

you should check the presence of low molecular weight bands on your gel (do you have a low molecular weight marker?)

You could use a small pore size nitrocellulose membrane (looks like it's better for small molecules)

you can check you did effectively transfer your proteins to the membrane with Ponceau S red staining (and try to adjust + or - the blotting time in order to transfer the protein completely without over-blotting)

then, if the antibody does not work with your standard conditions you can try to make it work... try to incubate it overnight at 4 °C with your membrane OR for 1-2 h at 37 °C...

Finally, if it does not work in western blot, it might still work in an indirect ELISA. (I assume you only have 1 antibody against your protein)... it is a very sensitive technique!

Good luck

By selected reaction monitoring (SRM).

SRM will require significant time investment to get it working so I would recommend testing some of the suggestions Leonardo described such as a Tricine gel before discarding the antibody approach altogether.

SRM might be a good alternative as it will be more sensitive. This would be dependant on the protein you are trying to detect. If it is small there may be limited peptides you might be able to detect, or use for quantitation if that is the ultimate aim. I have found small proteins, particularly secreted proteins (not sure if yours is), to be highly disulphide bonded and a little more complicated to deal with.

There are membranes with 0.1 micron pore size which might help (http://www.sigmaaldrich.com/labware/labware-products.html?TablePage=20469807) - this type is all that would work for our 5.3kDa protein, otherwise the protein has a tendency to "blow through" the membrane. If the protein is very hydrophilic/intrinsically disordered then this is more likely to be the case. 

http://www.sigmaaldrich.com/labware/labware-products.html?TablePage=20469807