Back

Protein Classifier for Thyroid Nodules Learned from Rapidly Acquired Proteotypes

Sun, Y.; Selvarajan, S.; Zang, Z.; Liu, W.; Zhu, Y. J.; Zhang, H.; Chen, H.; Cai, X.; Gao, H.; Wu, Z.; Chen, L.; Teng, X.; Zhao, Y.; Mantoo, S.; Lim, T. K.-H.; Hariraman, B.; Yeow, S.; Syed Abdillah, S. M. F.; Lee, S. S.; Ruan, G.; Zhang, Q.; Zhu, T.; Wang, W.; Wang, G.; Xiao, J.; He, Y.; Wang, Z.; Sun, W.; Qin, Y.; Xiao, Q.; Zheng, X.; Wang, L.; Zheng, X.; Xu, K.; Shao, Y.; Liu, K.; Zheng, S.; Li, S. Z.; Kon, O. L.; Iyer, N. G.; Guo, T.

2020-04-14 oncology
10.1101/2020.04.09.20059741 medRxiv
Show abstract

Up to 30% of thyroid nodules cannot be accurately classified as benign or malignant by cytopathology. Diagnostic accuracy can be improved by nucleic acid-based testing, yet a sizeable number of diagnostic thyroidectomies remains unavoidable. In order to develop a protein classifier for thyroid nodules, we analyzed the quantitative proteomes of 1,725 retrospective thyroid tissue samples from 578 patients using pressure-cycling technology and data-independent acquisition mass spectrometry. With artificial neural networks, a classifier of 14 proteins achieved over 93% accuracy in classifying malignant thyroid nodules. This classifier was validated in retrospective samples of 271 patients (91% accuracy), and prospective samples of 62 patients (88% accuracy) from four independent centers. These rapidly acquired proteotypes and artificial neural networks supported the establishment of an effective protein classifier for classifying thyroid nodules.

Matching journals

The top 10 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.