It's complicated.  There's no "best" tag for all situations.  If there was, then there wouldn't be multiple tags on offer.  A couple of things to consider: 

- What is the culture medium? 

- Does the protein have other exceptional characteristics, that could be used in conjunction with a tag?  (heat stability, strong charge, unusual size...?)

- How concentrated is your protein? 

- How large scale do you need to make it?  

- What kind of pre-processing do you have available (buffer exchange, concentrators...?)

- Do you need commercial use, and if yes, can you afford the royalties?  

- What's the intended use, and what requirements does that pose for residual protein and non-protein contaminants in your preparation...?

We'd need to be given much more background about what you're doing to come up with a coherent purification strategy.  And even then, the "best" answer can only be determined experimentally.  So I think the best "advice in a vacuum" here would be, prepare to try a number of variations of pre-processing, purification, post-processing steps, and analytic tests that tell you what you need to know about your preparation.  

All that said, His7, GST, MBP, Strep are all reasonable places to start.

It's complicated.  There's no "best" tag for all situations.  If there was, then there wouldn't be multiple tags on offer.  A couple of things to consider: 

- What is the culture medium? 

- Does the protein have other exceptional characteristics, that could be used in conjunction with a tag?  (heat stability, strong charge, unusual size...?)

- How concentrated is your protein? 

- How large scale do you need to make it?  

- What kind of pre-processing do you have available (buffer exchange, concentrators...?)

- Do you need commercial use, and if yes, can you afford the royalties?  

- What's the intended use, and what requirements does that pose for residual protein and non-protein contaminants in your preparation...?

We'd need to be given much more background about what you're doing to come up with a coherent purification strategy.  And even then, the "best" answer can only be determined experimentally.  So I think the best "advice in a vacuum" here would be, prepare to try a number of variations of pre-processing, purification, post-processing steps, and analytic tests that tell you what you need to know about your preparation.  

All that said, His7, GST, MBP, Strep are all reasonable places to start.

I think His tags are the best because they are easiest and purification is the fastest, simplest, and cheapest. The only reason to use fusion proteins like GST and MBP is if your protein is unstable or doesn't fold properly, ie. Comes out in inclusion bodies. You can put something like a TEV cleavage sequence between the tag and the protein, if you don't want it there.

Flag tag has a good binding specific. If large tag is acceptable, FC tag may be a good choice.

Problem also depends on the scale of your purification.

For large scale the above mentioned tags (like His tag) are fine. For small scale purifications (out of some cell culture dishes) it also works to purifiy proteins via immunoprecipitation with monoclonal antibodies for which the epitop is known. Example: express FLAG- or HA-tagged protein, perform immunoprecipitation and elute with FLAG or HA epitop peptide from the beads. ALternatively you could add a protease cleavage site (eg Precission or TEV) between tag and your protein and elute it from your beads after immunoprecipitation by cleavage with the protease.

What experience do people have with the Halo-tag system from Promega?

Their web protocols look quite impressive...