Biology subjects

Mendenhall, J.

Publications and source records attributed to Mendenhall, J..

3 recordsLinked to original sources

bioRxiv2020

A multi-task deep-learning system for predicting membrane associations and secondary structures of proteins

Accurate prediction of secondary structures and transmembrane segments is often the first step towards modeling the tertiary structure of a protein. Existing methods are either specialized in one class of proteins or developed to predict one type of 1D structural attributes (secondary structure, topology, or transmembrane segment). In this work, we develop a new method for simultaneous prediction of secondary structure, transmembrane segment, and transmembrane topology with no a priori assumption on the class of the input protein sequence. The new method, Membrane Association and Secondary Structures of Proteins (MASSP) predictor, uses multi-tiered neural networks that incorporate recent innovations in machine learning. The first tier is a multi-task multi-layer convolutional neural network (CNN) that learns patterns in image-like input position-specific-scoring matrices (PSSMs) and predicts residue-level 1D structural attributes. The second tier is a long short-term memory (LSTM) neural network that treats the predictions of the first tier from the perspective of natural language processing and predicts the class of the input protein sequence. We curated a non-redundant data set consisting of 54 bitopic, 241 multi-spanning TM-alpha, 77 TM-beta, and 372 soluble proteins, respectively for training and testing MASSP. For secondary structure prediction, the mean three-state accuracy (Q3) of MASSP is 0.830, better than the Q3 of PSIPRED (0.829) and that of SPINE-X (0.813) and substantially better than that of Jufo9D (0.762) and RaptorX-Property (0.741). The mean segment overlap score (SOV) of MASSP is 0.752, gaining at least 7.7% improvement over all the other four methods. For transmembrane topology prediction, MASSP has a performance comparable to OCTOPUS and substantially better than MEMSAT3 and TMHMM2 on TM-alpha proteins, and on TM-beta proteins, MASSP is significantly better than both BOCTOPUS2 and PRED-TMBB2. By integrating prediction of secondary structure and transmembrane segments in a deep-learning framework, MASSP improves performance over previous methods, has broader applicability, and enables proteome scale predictions.

bioinformatics↗

bioRxiv2020

Deep Learning for Virtual Screening: Five Reasons to Use ROC Cost Functions

Computer-aided drug discovery is an essential component of modern drug development. Therein, deep learning has become an important tool for rapid screening of billions of molecules in silico for potential hits containing desired chemical features. Despite its importance, substantial challenges persist in training these models, such as severe class imbalance, high decision thresholds, and lack of ground truth labels in some datasets. In this work we argue in favor of directly optimizing the receiver operating characteristic (ROC) in such cases, due to its robustness to class imbalance, its ability to compromise over different decision thresholds, certain freedom to influence the relative weights in this compromise, fidelity to typical benchmarking measures, and equivalence to positive/unlabeled learning. We also propose new training schemes (coherent mini-batch arrangement, and usage of out-of-batch samples) for cost functions based on the ROC, as well as a cost function based on the logAUC metric that facilitates early enrichment (i.e. improves performance at high decision thresholds, as often desired when synthesizing predicted hit compounds). We demonstrate that these approaches outperform standard deep learning approaches on a series of PubChem high-throughput screening datasets that represent realistic and diverse drug discovery campaigns on major drug target families.

pharmacology and toxicology↗

bioRxiv2019

Deep Learning-Based Point-Scanning Super-Resolution Imaging

Point scanning imaging systems (e.g. scanning electron or laser scanning confocal microscopes) are perhaps the most widely used tools for high resolution cellular and tissue imaging. Like all other imaging modalities, the resolution, speed, sample preservation, and signal-to-noise ratio (SNR) of point scanning systems are difficult to optimize simultaneously. In particular, point scanning systems are uniquely constrained by an inverse relationship between imaging speed and pixel resolution. Here we show these limitations can be mitigated via the use of deep learning-based super-sampling of undersampled images acquired on a point-scanning system, which we termed point-scanning super-resolution (PSSR) imaging. Oversampled, high SNR ground truth images acquired on scanning electron or Airyscan laser scanning confocal microscopes were crappified to generate semi-synthetic training data for PSSR models that were then used to restore real-world undersampled images. Remarkably, our EM PSSR model could restore undersampled images acquired with different optics, detectors, samples, or sample preparation methods in other labs. PSSR enabled previously unattainable 2 nm resolution images with our serial block face scanning electron microscope system. For fluorescence, we show that undersampled confocal images combined with a multiframe PSSR model trained on Airyscan timelapses facilitates Airyscan-equivalent spatial resolution and SNR with [~]100x lower laser dose and 16x higher frame rates than corresponding high-resolution acquisitions. In conclusion, PSSR facilitates point-scanning image acquisition with otherwise unattainable resolution, speed, and sensitivity.\n\n\n\nO_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=116 SRC=\"FIGDIR/small/740548v8_ufig1.gif\" ALT=\"Figure 1\">\nView larger version (90K):\norg.highwire.dtl.DTLVardef@fbf19aorg.highwire.dtl.DTLVardef@b2cc88org.highwire.dtl.DTLVardef@921e1corg.highwire.dtl.DTLVardef@f6c5af_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioinformatics↗