When compiling a list of prospective Twitter leads for any subject area, an obvious approach is to find accounts that tweet about that subject. This can be done with the streaming API or REST API. This still leaves a lot of curating work to select the highest quality accounts from this group.
A valuable shortcut is to build on the selection work already done by experts on that subject. When you identify an account that speaks authoritatively and focuses largely on your topic, you can copy the accounts they follow into a database, and use that as the starting point for a good lead list.
The list of accounts an expert is following will almost always be smaller than those who are following them, but the quality will be much higher. Spam followers are a chronic problem, and any list of followers will have a high number of spam and bot accounts. Nobody will deliberately follow a spam account, so the people the expert has chosen to follow is going to be much better.
A following list is not going to be perfect, however. Twitter users are human and have other interests, so you still have to review the accounts they are following to filter out inappropriate accounts. There is also the long-term Twitter bug that has most accounts randomly follow others at a low rate. All the accounts I manage seem to follow at least 10-20 accounts a day all by themselves. They are pretty easy to spot, since they are often in languages I don’t read.
From a technical viewpoint, you need to use the /friends/ids API call to get the accounts a user is following. You can find a complete PHP script for using this API to collect friends in my engagement programming source code.