This is the seventh state of bims statement (sobs). Sobs comes out every 30 of January. It commemorates the first sober meeting between Gavin and Thomas on 30th January 2017. This statement contains a part by Gavin and a part by Thomas.

Gavin writes:

The past year for Biomed News has seen a modest increase in selectors, larger numbers of subscribers and other activity.

At the time of writing there are 107 active reports, an approximate increase of 20 reports in the past year. This slow uptake is still frustrating. Our report topics remain centred on mitochondria, metabolism, cancer and immunology. Topics have marginally become broader. More selectors are now managing more than one report.

Issues of our reports continue to be shared. We have 2056 followers on X, 235 followers on LinkedIn, and 387 on Blue Sky. With the demise of X for science users, Blue Sky has taken over in some aspects for the scientific community. Recruitment of selectors through these three platforms continues. Blue Sky community seems receptive so far to requests to start Biomed News reports.

I have presented about Biomed News at the Edinburgh Open Research Conference in May and the Liverpool John Moores University research conference in December. There was again some general interest but nothing came from it. I got into discussions with MitoWorld, an online mitochondrial community platform, about the potential use of Biomed News reports as a way to inform the mitochondrial community of the latest literature.

Thomas writes:

Gavin used to run a Google form to open reports. He ran into some problem with this and requested we should do our own. I took on the job. This is what I worked on at the start of the year. The aspect I was most worried about was the entry of correct PubMed identifiers. Interestingly enough, developing a background call to PubMed to validate numbers entered by users was not that much work. And in fact users manage to enter PubMed numbers correctly. But the interaction on checking form and resubmission took me weeks to do. Basically, what seems to more trivial task took longest to do.

In May the mirroring software for PubMed stopped working, so I had write new software in a haste. Without a timely download of PubMed data, we are in big trouble.

Among a flurry of opened reports that did not go beyond the first issue in the Summer, Gavin and I examined the initial proposed issue served to budding selectors. We found that they were badly presorted. I need to improve. I rewrote the entire code base in this of early-stage reports. These are now defined as reports that are less than two years old. That took about 10 weeks full-time. The rewrite seems to be at least partly sucessful. Gavin now advises that while debut issues are not good, they are no-nonsense and usable.

The machine-learning code rewrite also I introduced the sci-spacy software from the Allan institute. The number of multi-word features rose as a result. I believe this let to an improvement in presorting for all reports.

Happening a bit seperately but at the same time, I introduced report forking. This is a new method to train a starting report. It is alternative to seeding. It is available when there a parent report that is close the child report. Seeding uses a limited number of issues from the parent report, say ten to illustrate. The dates of these issues do not change through the process. At the start, training selections for the parent and child report are the same. After the first issue, training for the child report uses ten issues from the parent report and one issue from the child report. After the second issue, training for the child report uses nine issues from the parent and two from the child and so on. After eleven issues, the child report only uses its own training data.

Finally as part of the broader reorganization I set a time limit for data from issues to expire. At this time it is set at 100 issuedates, so basically 100 weeks. This imposes a limit on the train of very old reports. I will need to go back to this.

In October and November, I worked on back issues. Now we have a page for each issue, at the convenient URL https://biomed.news/yyyy-mm-dd. These pages demonstrate my tender loving care for historic data. For another example, I fully support affiliation histories. Bims mentions the institution a selector worked at at the time of the issue.

While I am still actively developing the system, I also have to constantly fight various fires. Let me give some example from the end of last year. It was particularly problem-prone. First, Gavin found that when selectors caught up with reports, ernad sometimes skipped the last issues. I hope I have fixed this bug. I created these missing issues and informed the affected selectors. They took the news in good spirit. Second, in the early hours 2025‒01‒05 release, Marcus Karlstetter reported dozens of papers with the same title in his proposed issue. The problem came directly from PubMed. I assumed that PubMed would fix the records. I wrote to them immediately but surely they would come back on Monday morning only. In the meantime I rushed to write software to assess the scale of the problem. I informed selectors that had the affected papers in their proposed issues. PubMed came back to me in very timely fashion. It turns out that the records were dumped by Wiley. PubMed considered them out of scope. They would be deleted. Thus the only reasonable course of action was for users to accept relevant record at pickup and remove them at filtering. Third, PubMed ran their annual update, and I had to increment my software written for downloading that I wrote in May. When that was over an attacker tried various ways to probe security flaws on the server. I did some further changes to ernad to return the login screen when doggy input is given.

Both write:

Once again, we thank you for your support and the time you spend to keep your reports up to date and your comments to improve the system. Wishing everyone a successful 2025.