“… a false statement of fact, made deliberately, is the most serious crime a scientist can commit.” English Chemist and novelist, Charles Percy Snow (1905-1980)
Over the last few years, there has been a growing number of papers that have been retracted due to falsified results. This takes shape in the form of doctored figures. Biological images such as PCRs and Western Blots.
But is this such a problem you ask? Well, results from Western Blots and PCRs underpin biomedical research. Expensive, time exhausting, hard work. However, when scientists start to edit and alter their images in order to get publications and grants, well... that's truly atrocious. For many reasons, not just because it's fraud and the ethics behind it.
We have personally known postdocs and students who have wasted many months and years chasing things results and trying to replicate experience which turns out not to be valid.
So we must stop this ASAP!
How you ask?
We have spoken to numerous colleagues. We have done an exhaustive search, and sadly there isn't a tool like this that exists specifically for biological images! That is when we decided we must take this into our own hands. We contacted our scientific colleagues and gathered as many figures as possible. Thousands upon thousands globally. We also mined images manually from papers online and also known retracted papers which were retracted due to editing images.
What it does
Our tool uses machine learning algorithms to detect altered biological figures. Those images which have had parts of a western blot band edited or histological stains edited/duplicated/doctored are picked up.
How we built it
We first had to gather a lot of information! Lots of images, thousands! We used MATLAB to build this tool.
Challenges we ran into
Obtaining lots of PCR, Western Blots, and Histological images. We are both beginners in coding, in particular MATLAB, and therefore we really struggled to debug. Also lots of technical issues, RE Wifi!! A hackathon is hard doing it from home!
Accomplishments that we're proud of
We have built something we really care about and passionate about. We are really proud of how far it manage to get it within 24 hours!
What we learned
Coding skills! Team skills - good communication, delegating tasks to each other and knowing when to ask for help. We also learnt a lot from the talks and workshops put on by HackMed, so thank you for that!
What's next for Retraction Action!!
Thanks to this Hackathon we have actually been approached by a research group that wants to help us develop this tool further. So keep your eyes open!!