Great question!
The interface in DirectAdmin (Spamassassin Setup) is used to write to the user_prefs file, which is then translated into a user preference file for rspamd. Any manual configurations outside of what is there would not translate properly to rspamd. For example, wildcards are removed in translation for rspamd as they are not supported.
No additional manual configuration of spam filters would function at the user level. However, if you wish to contribute to our rspamd inbound filters, you can find pieces of it in the repositories here:
https://git.mxroute.com/jarland
Examples:
Spammy subjects - https://git.mxroute.com/jarland/rspamd_locald/src/branch/master/spammy_subject.map
Spammy From headers - https://git.mxroute.com/jarland/rspamd_locald/src/branch/master/spammy_from.map
(This uses regex)
Accepted pull requests are pushed to the servers every hour. The idea behind these filters are to focus purely on common trends that deserve outright rejection.
You can also use sieve filters to gain more control over your email. This is either accomplished by an external client over port 4190, or through Roundcube under Settings > Filters.