First of all, Leica Microsystems would like to thank all 2,172 teams which participated for their hard work, creative ideas, and contributions. Leica Microsystems would also like to congratulate the winners of the prize money.
The top 4 prize-winning teams with the highest F1 scores are:
- 1st place → team “bestfitting” ($14,000 prize);
- 2nd place → team “WAIR” ($10,000 prize);
- 3rd place → team “pudae” ($8,000 prize);
- 4th place → team “Wienerschnitzelgemeinschaft” ($5,000 prize).
The objective of each team participating in the competition was to build a predictive model using artificial intelligence (AI) and data mining methods to classify 28 types of cell organelles for protein localization. More than 170,000 confocal microscope images of cultured cells were made available by the Human Protein Atlas (HPA). Many of these images were acquired with Leica confocal instruments. The image data were analyzed by the models developed during the competition and the results evaluated to determine the winners. The competition organizers were amazed by the overall participation and the achieved results.
Life scientists study the localization of proteins at cell organelles to better understand their role in health and disease. By automating this task via the exploitation of AI and data mining, scientists will be able to detect abnormalities in the cell more rapidly and accurately. Faster, more precise predictive models will allow scientists to put more focus on rare cell organelles and phenotypes which, at present, often go overlooked.
To learn more about the HPA and protein localization, you can read an interview with Emma Lundberg. She is the lead scientist behind the Cell Atlas, a part of the HPA, and the SciLifeLab at the KTH Royal Institute of Technology in Stockholm, Sweden.