Biology subjects

Frolov, D.

Publications and source records attributed to Frolov, D..

3 recordsLinked to original sources

bioRxiv2025

Inferring binding specificities of human transcription factors with the wisdom of crowds

DNA motif discovery and, particularly, computational modeling of transcription factor binding motifs, has been a mecca of algorithmic bioinformatics for several decades. Here, we report the results of the largest open community challenge in Inferring BInding Specificities (IBIS), where participants all over the world were invited to construct binding specificity models from multi-assay experimental data for poorly studied human transcription factors. The submissions were rigorously tested against a rich held-out dataset. Benchmarking demonstrated a consistent advantage of properly designed deep learning models over traditional positional weight matrices and other machine learning methods. Yet, the positional weight matrices displayed a surprisingly strong performance out of the box, being only slightly behind the best deep learning models. A post-challenge assessment of a selection of other deep learning methods further solidified this finding. IBIS highlights the power of benchmarking in finding adequate DNA motif representations, emphasizes the pros and cons of various machine learning methods applied to DNA motif modeling, and establishes a rich dataset, benchmarking protocols, and computational framework for a fair cross-platform evaluation of future models of transcription factor binding motifs in DNA sequences. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=175 SRC="FIGDIR/small/688692v1_ufig1.gif" ALT="Figure 1"> View larger version (64K): org.highwire.dtl.DTLVardef@1c6677corg.highwire.dtl.DTLVardef@b4124aorg.highwire.dtl.DTLVardef@1ce2b1org.highwire.dtl.DTLVardef@66e917_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioinformatics↗

bioRxiv2024

Quantum chemical properties of chlorinated polycyclic aromatic hydrocarbons for delta machine learning

Promising {Delta}-machine learning approaches aim to correct the values of molecular properties obtained with computationally inexpensive methods to the accuracy of higher levels of theory. Training such models requires datasets containing the results of calculations at several different levels of quantum chemical theory. While several large and chemically diverse datasets have been published, studies in many areas require specialized datasets of structurally related molecules. Chlorinated polycyclic aromatic hydrocarbons (Cl-PAHs), the products of incomplete combustion of organic substances and materials, are hazardous pollutants with carcinogenic and mutagenic activity. Quantum chemistry methods are important to understand their formation mechanisms and properties. We describe a dataset, PACHQA, containing the results of quantum chemical calculations including properties, geometries, wavefunctions, and electron densities for 3551 molecules including 3417 Cl-PAHs with up to 6 rings and a different number of chlorine atoms in their structure as well as 134 parent polycyclic aromatic hydrocarbons (PAHs). The major part of these molecules have previously not been included in any quantum chemical datasets. The calculations were performed at three different levels of theory including geometry optimization with GFN2-xTB and r2SCAN-3c methods and single-point energy calculation at {omega}B97X-D4/def2-TZVP DFT level. The dataset can be useful to develop and validate the computational, machine learning, or experimental, approaches and study the structure-property relationships for Cl-PAHs.

biophysics↗

bioRxiv2022

Single Pixel Reconstruction Imaging: taking confocal imaging to the extreme

Light nanoscopy is attracting widespread interest for the visualization of fluorescent structures at the nanometer scale, especially in cellular biology. To achieve nanoscale resolution, one has to surpass the diffraction limit--a fundamental phenomenon determining the spot size of focused light. Recently, a variety of methods have overcome this limit, yet in practice they are often constrained by the requirement of special fluorophores, nontrivial data processing, or high price and complex implementation. For this reason, confocal fluorescence microscopy that yields relatively low resolution is still the dominant method in biomedical sciences. It was shown that image scanning microscopy (ISM) with an array detector instead of a point detector could improve the resolution of confocal microscopy. Here we review the principles of the confocal microscopy and present a simple method based on ISM with a different image reconstruction approach, which can be easily implemented in any camera-based laser-scanning set-up to experimentally obtain the theoretical resolution limit of the confocal microscopy. Our method, Single Pixel Reconstruction Imaging (SPiRI) enables high-resolution 3D imaging utilizing image formation only from a single pixel of each of the recorded frames. We achieve experimental axial resolution of 330 nm, which was not shown before by basic confocal or ISM-based systems. Contrary to the majority of techniques, SPiRI method exhibits a low lateral-to-axial FWHM aspect ratio, which means a considerable improvement in 3D fluorescence imaging of cellular structures. As a demonstration of SPiRI application in biomedical sciences, we present a 3D structure of bacterial chromosome with excellent precision.

biophysics↗