As with any website attack you need to not only remove the malicious code as others have stated above, but you must also find *how* the attacker was able to inject the code/file in the first place.
Once you have detected the attack request, it will give you some clue as to what part of your website is vulnerable to such an attack.
If there is no evidence in the log files that disclose how an attacker was able to add, prepend, append code or files onto your site then you need to look at the webserver itself.
This is where it gets more difficult because most webhosts will not disclose that they have patched a vulnerable server, or admit that their servers are insecure. One way to tell is to google other websites on your server to see if they are also complaining of being attacked. If so then it may well be a server wide issue which you cannot fix yourself other than moving to a more secure webservice.