Estimating the difference between structure-factor amplitudes using multivariate Bayesian inference

Modeling for predicting survival fraction of cells after ultra-high dose rate irradiation

Objective. FLASH radiotherapy (FLASH-RT) with ultra-high dose rate (UHDR) irradiation (i.e. > 40 Gy s-1) spares the function of normal tissues while preserving antitumor efficacy, known as the FLASH effect. The biological effects after conventional dose rate-radiotherapy (CONV-RT) with ≤0.1 Gy s-1have been well modeled by considering microdosimetry and DNA repair processes, meanwhile modeling of radiosensitivities under UHDR irradiation is insufficient. Here, we developed anintegrated microdosimetric-kinetic(IMK)model for UHDR-irradiationenabling the prediction of surviving fraction after UHDR irradiation.Approach.TheIMK model for UHDR-irradiationconsiders the initial DNA damage yields by the modification of indirect effects under UHDR compared to CONV dose rate. The developed model is based on the linear-quadratic (LQ) nature with the dose and dose square coefficients, considering the reduction of DNA damage yields as a function of dose rate.Main results.The estimate by the developed model could successfully reproduce thein vitroexperimental dose-response curve for various cell line types and dose rates.Significance.The developed model would be useful for predicting the biological effects under the UHDR irradiation.

Estimating the probability of coincidental similarity between atomic displacement parameters with machine learning

High-resolution diffraction studies of macromolecules incorporate the tensor form of the anisotropic displacement parameter (ADP) of atoms from their mean position. The comparison of these parameters requires a statistical framework that can handle the experimental and modeling errors linked to structure determination. Here, a Bayesian machine learning model is introduced that approximates ADPs with the random Wishart distribution. This model allows for the comparison of random samples from a distribution that is trained on experimental structures. The comparison revealed that the experimental similarity between atoms is larger than predicted by the random model for a substantial fraction of the comparisons. Different metrics between ADPs were evaluated and categorized based on how useful they are at detecting non-accidental similarity and whether they can be replaced by other metrics. The most complementary comparisons were provided by Euclidean, Riemann and Wasserstein metrics. The analysis of ADP similarity and the positional distance of atoms in bovine trypsin revealed a set of atoms with striking ADP similarity over a long physical distance, and generally the physical distance between atoms and their ADP similarity do not correlate strongly. A substantial fraction of long- and short-range ADP similarities does not form by coincidence and are reproducibly observed in different crystal structures of the same protein.

Bayesian machine learning improves single-wavelength anomalous diffraction phasing

Single-wavelength X-ray anomalous diffraction (SAD) is a frequently employed technique to solve the phase problem in X-ray crystallography. The precision and accuracy of recovered anomalous differences are crucial for determining the correct phases. Continuous rotation (CR) and inverse-beam geometry (IBG) anomalous data collection methods have been performed on tetragonal lysozyme and monoclinic survivin crystals and analysis carried out of how correlated the pairs of Friedel's reflections are after scaling. A multivariate Bayesian model for estimating anomalous differences was tested, which takes into account the correlation between pairs of intensity observations and incorporates the a priori knowledge about the positivity of intensity. The CR and IBG data collection methods resulted in positive correlation between I(+) and I(-) observations, indicating that the anomalous difference dominates between these observations, rather than different levels of radiation damage. An alternative pairing method based on near simultaneously observed Bijvoet's pairs displayed lower correlation and it was unsuccessful for recovering useful anomalous differences when using the multivariate Bayesian model. In contrast, multivariate Bayesian treatment of Friedel's pairs improved the initial phasing of the two tested crystal systems and the two data collection methods.

Predicting Protein-Protein Interactions from Matrix-Based Protein Sequence Using Convolution Neural Network and Feature-Selective Rotation Forest

Protein is an essential component of the living organism. The prediction of protein-protein interactions (PPIs) has important implications for understanding the behavioral processes of life, preventing diseases, and developing new drugs. Although the development of high-throughput technology makes it possible to identify PPIs in large-scale biological experiments, it restricts the extensive use of experimental methods due to the constraints of time, cost, false positive rate and other conditions. Therefore, there is an urgent need for computational methods as a supplement to experimental methods to predict PPIs rapidly and accurately. In this paper, we propose a novel approach, namely CNN-FSRF, for predicting PPIs based on protein sequence by combining deep learning Convolution Neural Network (CNN) with Feature-Selective Rotation Forest (FSRF). The proposed method firstly converts the protein sequence into the Position-Specific Scoring Matrix (PSSM) containing biological evolution information, then uses CNN to objectively and efficiently extracts the deeply hidden features of the protein, and finally removes the redundant noise information by FSRF and gives the accurate prediction results. When performed on the PPIs datasets Yeast and Helicobacter pylori, CNN-FSRF achieved a prediction accuracy of 97.75% and 88.96%. To further evaluate the prediction performance, we compared CNN-FSRF with SVM and other existing methods. In addition, we also verified the performance of CNN-FSRF on independent datasets. Excellent experimental results indicate that CNN-FSRF can be used as a useful complement to biological experiments to identify protein interactions.

Asymmetry in serial femtosecond crystallography data

Serial crystallography is an increasingly important approach to protein crystallography that exploits both X-ray free-electron laser (XFEL) and synchrotron radiation. Serial crystallography recovers complete X-ray diffraction data by processing and merging diffraction images from thousands of randomly oriented non-uniform microcrystals, of which all observations are partial Bragg reflections. Random fluctuations in the XFEL pulse energy spectrum, variations in the size and shape of microcrystals, integrating over millions of weak partial observations and instabilities in the XFEL beam position lead to new types of experimental errors. The quality of Bragg intensity estimates deriving from serial crystallography is therefore contingent upon assumptions made while modeling these data. Here it is observed that serial femtosecond crystallography (SFX) Bragg reflections do not follow a unimodal Gaussian distribution and it is recommended that an idealized assumption of single Gaussian peak profiles be relaxed to incorporate apparent asymmetries when processing SFX data. The phenomenon is illustrated by re-analyzing data collected from microcrystals of the Blastochloris viridis photosynthetic reaction center and comparing these intensity observations with conventional synchrotron data. The results show that skewness in the SFX observations captures the essence of the Wilson plot and an empirical treatment is suggested that can help to separate the diffraction Bragg intensity from the background.