One idea: You can play with the github dataset in ClickHouse playground.

This is just a quick sql query to look for the DOI number pattern mentioned in any comment on repositories:

https://play.clickhouse.com/play?user=play#c2VsZWN0IHJlcG9fb...

``` select repo_name,event_type,body from github_events where event_type in ('IssueCommentEvent','IssuesEvent','PullRequestEvent','PullRequestReviewCommentEvent') and match (body,'.10\.\d{4,9}\/[-\._;()\/:A-Z0-9]+.') limit 10 ```

perhaps you can extend on that :)