RESUMEN
Complex mixtures containing natural products are still an interesting source of novel drug candidates. High content screening (HCS) is a popular tool to screen for such. In particular, multiplexed HCS assays promise comprehensive bioactivity profiles, but generate also high amounts of data. Yet, only some machine learning (ML) applications for data analysis are available and these usually require a profound knowledge of the underlying cell biology. Unfortunately, there are no applications that simply predict if samples are biologically active or not (any kind of bioactivity). Within this work, we benchmark ML algorithms for binary classification, starting with classical ML models, which are the standard classifiers of the scikit-learn library or ensemble models of these classifiers (a total of 92 models tested). Followed by a partial least square regression (PLSR)-based classification (44 tested models in total) and simple artificial neural networks (ANNs) with dense layers (72 tested models in total). In addition, a novelty detection (ND) was examined, which is supposed to handle unknown patterns. For the final analysis the models, with and without upstream ND, were tested with two independent data sets. In our analysis, a stacking model, an ensamble model of class ML algorithms, performed best to predict new and unknown data. ND improved the predictions of the models and was useful to handle unknown patterns. Importantly, the classifier presented here can be easily rebuilt and be adapted to the data and demands of other groups. The hit detector (ND + stacking model) is universal and suitable for a broader application to support the search for new drug candidates.
RESUMEN
Atom probe tomography (APT) can theoretically deliver accurate chemical and isotopic analyses at a high level of sensitivity, precision, and spatial resolution. However, empirical APT data often contain significant biases that lead to erroneous chemical concentration and isotopic abundance measurements. The present study explores the accuracy of quantitative isotopic analyses performed via atom probe mass spectrometry. A machine learning-based adaptive peak fitting algorithm was developed to provide a reproducible and mathematically defensible means to determine peak shapes and intensities in the mass spectrum for specific ion species. The isotopic abundance measurements made with the atom probe are compared directly with the known isotopic abundance values for each of the materials. Even in the presence of exceedingly high numbers of multi-hit detection events (up to 80%), and in the absence of any deadtime corrections, our approach produced isotopic abundance measurements having an accuracy consistent with values limited predominantly by counting statistics.
RESUMEN
Serial femtosecond crystallography is an emerging and promising method for determining protein structures, making use of the ultrafast and bright X-ray pulses from X-ray free-electron lasers. The upcoming X-ray laser sources will produce well above 1000â pulses per second and will pose a new challenge: how to quickly determine successful crystal hits and avoid a high-rate data deluge. Proposed here is a hit-finding scheme based on detecting photons from plasma emission after the sample has been intercepted by the X-ray laser. Plasma emission spectra are simulated for systems exposed to high-intensity femtosecond pulses, for both protein crystals and the liquid carrier systems that are used for sample delivery. The thermal radiation from the glowing plasma gives a strong background in the XUV region that depends on the intensity of the pulse, around the emission lines from light elements (carbon, nitrogen, oxygen). Sample hits can be reliably distinguished from the carrier liquid based on the characteristic emission lines from heavier elements present only in the sample, such as sulfur. For buffer systems with sulfur present, selenomethionine substitution is suggested, where the selenium emission lines could be used both as an indication of a hit and as an aid in phasing and structural reconstruction of the protein.