David B. Thomas

Protein interactions are crucial for understanding biological functions and disease mechanisms, but predicting these remains a complex task in computational biology. Increasingly, Deep Learning models are having success in interface prediction. This study presents PIPENN-EMB which explores the added value of using embeddings from the ProtT5-XL protein language model. Our results show substantial improvement over the previously published PIPENN model for protein interaction interface prediction, reaching an MCC of 0.313 vs. 0.249, and AUROC 0.800 vs. 0.755 on the BIO_DL_TE test set. We furthermore show that these embeddings cover a broad range of ‘hand-crafted’ protein features in ablation studies. PIPENN-EMB reaches state-of-the-art performance on the ZK448 dataset for protein-protein interface prediction. We showcase predictions on 25 resistance-related proteins from Mycobacterium tuberculosis. Furthermore, whereas other state-of-the-art sequence-based methods perform worse for proteins that have little recognisable homology in their training data, PIPENN-EMB generalises to remote homologs, yielding stable AUROC across all three test sets with less than 30% sequence identity to the training dataset, and even to proteins with less than 15% sequence identity.

Protein interactions are crucial for understanding biological functions and disease mechanisms, but predicting these remains a complex task in computational biology. Increasingly, Deep Learning models are having success in interface prediction. This study presents PIPENN-EMB which explores the added value of using embeddings from the ProtT5-XL protein language model. Our results show substantial improvement over the previously published PIPENN model for protein interaction interface prediction, reaching an MCC of 0.313 vs. 0.249, and AUC-ROC 0.800 vs. 0.755 on the BIO_DL_TE test set. We furthermore show that these embeddings cover a broad range of ‘hand-crafted’ protein features in ablation studies. PIPENN-EMB reaches state-of-the-art performance on the ZK448 dataset for protein-protein interface prediction. We showcase predictions on 25 resistance-related proteins from Mycobacterium tuberculosis . Furthermore, whereas other state-of-the-art sequence-based methods perform worse for proteins that have little recognisable homology in their training data, PIPENN-EMB generalises to remote homologs, yielding stable AUC-ROC across all three test sets with less than 30% sequence identity to the training dataset, and even to proteins with less than 15% sequence identity. Availability: Webserver, source code and datasets at www.ibi.vu.nl/programs/pipennemb/

Objective: A case‐cohort study was conducted to investigate associations between occupational exposures and endometrial cancer nested within a large cohort of textile workers in Shanghai, China.Methods: The study included 176 incident endometrial cancer cases diagnosed from 1989 to 1998 and a randomly‐selected age‐stratified reference subcohort (n = 3,061). Study subjects' complete work histories were linked to a job‐exposure matrix developed specifically for the textile industry to assess occupational exposures. Hazard ratios (HR) and 95% confidence intervals were calculated using Cox proportional hazards modeling adapted for the case‐cohort design, adjusting for age at menarche and a composite variable of gravidity and parity.Results: An increased risk of endometrial cancer was detected among women who had worked for ≥10 years in silk production (HR = 3.8, 95% CI 1.2–11.8) and had exposure to silk dust (HR = 1.7, 95% CI 0.9–3.4). Albeit with few exposed women (two cases and eight subcohort women), there was a 7.4‐fold increased risk associated with ≥10 years of silica dust exposure (95% CI 1.4–39.7).Conclusions: The findings suggest that some textile industry exposures might play a role in endometrial carcinoma and should be further replicated in other occupational settings. Am. J. Ind. Med. 51:673–679, 2008. Published 2008 Wiley‐Liss, Inc.

Breast cancer is commonly diagnosed at late stages in countries with limited resources. Efforts aimed at early detection can reduce the stage at diagnosis, potentially improving the odds o...