Background: Over the last few decades, a search for the theory of protein folding has grown into a full-fledged research field at the intersection of biology, chemistry and informatics. Despite enormous effort, there are still open questions and challenges, like understanding the rules by which amino acid sequence determines protein secondary structure. Objective: In this review, we depict the progress of the prediction methods over the years and identify sources of improvement. Methods: The protein secondary structure prediction problem is described followed by the discussion on theoretical limitations, description of the commonly used data sets, features and a review of three generations of methods with the focus on the most recent advances. Additionally, methods with available online servers are assessed on the independent data set. Results: The state-of-the-art methods are currently reaching almost 88% for 3-class prediction and 76.5% for an 8-class prediction. Conclusion: This review summarizes recent advances and outlines further research directions.

BackgroundSolitary pink lesions in differential diagnosis with hypopigmented/amelanotic melanoma present a diagnostic challenge in daily practice and are regularly referred for second expert opinion. Reflectance confocal microscopy (RCM) has been shown to improve diagnostic accuracy of dermoscopically equivocal pink lesions. No studies have been performed to evaluate the effect of adding a second expert reader and automatic removal of lesions with discordant management recommendations and its potential effect on diagnostic sensitivity and final management of these lesions in retrospective or telemedicine settings.ObjectiveTo improve diagnostic accuracy and reduce potential mismanagement of dermoscopically equivocal pink cutaneous lesions by implementing double reader concordance evaluation of RCM images.Methods316 dermoscopically equivocal pink lesions with dermoscopy-RCM image sets were evaluated retrospectively. Accuracy of three readers was evaluated by single reader evaluation of dermoscopy only and dermoscopy-RCM image sets and finally by double reader evaluation of dermoscopy-RCM image sets. Lesions with discordant diagnosis between two readers were automatically recommended for excision.ResultsDermoscopy only evaluation resulted in an overall sensitivity of 95.9% and specificity of 33.6%, with 1 of 12 amelanotic melanomas mismanaged. Dermoscopy-RCM image set single reader evaluation resulted in an overall sensitivity of 93.9% and overall specificity of 54.2%, with 1 of 12 melanomas mismanaged. Dermoscopy-RCM image set double reader concordance evaluation resulted in an overall sensitivity of 98.3% and specificity of 42.7%, with no amelanotic melanoma mismanagement.ConclusionEvaluation of dermoscopy-RCM image sets of equivocal pink lesions by a single reader in telemedicine settings is limited by the potential for misdiagnosis of dangerous malignant lesions. Double reader concordance evaluation with automatic referral of lesions for removal in the case of discordant diagnosis improves the diagnostic sensitivity in this subset of lesions and reduce potential misdiagnosis in settings where a second expert opinion may be employed.

No abstract

No abstract

Predicting protein function and structure from sequence remains an unsolved problem in bioinformatics. The best performing methods rely heavily on evolutionary information from multiple sequence alignments, which means their accuracy deteriorates for sequences with a few homologs, and given the increasing sequence database sizes requires long computation times. Here, a single-sequence-based prediction method is presented, called ProteinUnet, leveraging an U-Net convolutional network architecture. It is compared to SPIDER3-Single model, based on long short-term memory-bidirectional recurrent neural networks architecture. Both methods achieve similar results for prediction of secondary structures (both threeand eight-state), half-sphere exposure, and contact number, but ProteinUnet has two times fewer parameters, 17 times shorter inference time, and can be trained 11 times faster. Moreover, ProteinUnet tends to be better for short sequences and residues with a low number of local contacts. Additionally, the method of loss weighting is presented as an effective way of increasing accuracy for rare secondary structures.