ToxDL: deep learning using primary structure and domain embeddings for assessing protein toxicity.

MOTIVATION: Genetically engineering food crops involves introducing proteins from other species into crop plant species or modifying already existing proteins with gene editing techniques. In addition, newly synthesized proteins can be used as therapeutic protein drugs against diseases. For both research and safety regulation purposes, being able to assess the potential toxicity of newly introduced/synthesized proteins is of high importance. RESULTS: In this study, we present ToxDL, a deep learning-based approach for in silico prediction of protein toxicity from sequence alone. ToxDL consists of (i) a module encompassing a convolutional neural network that has been designed to handle variable-length input sequences, (ii) a domain2vec module for generating protein domain embeddings and (iii) an output module that classifies proteins as toxic or non-toxic, using the outputs of the two aforementioned modules. Independent test results obtained for animal proteins and cross-species transferability results obtained for bacteria proteins indicate that ToxDL outperforms traditional homology-based approaches and state-of-the-art machine-learning techniques. Furthermore, through visualizations based on saliency maps, we are able to verify that the proposed network learns known toxic motifs. Moreover, the saliency maps allow for directed in silico modification of a sequence, thus making it possible to alter its predicted protein toxicity. AVAILABILITY AND IMPLEMENTATION: ToxDL is freely available at http://www.csbio.sjtu.edu.cn/bioinf/ToxDL/. The source code can be found at https://github.com/xypan1232/ToxDL. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

A mixed film composed of oligonucleotides and poly(2-hydroxyethyl methacrylate) brushes to enhance selectivity for detection of single nucleotide polymorphisms.

Preliminary studies of mixed films composed of oligonucleotides and poly(2-hydroxyethyl methacrylate) (PHEMA) have recently been shown to enhance the selectivity for detection of 3 base-pair mismatched (3 bpm) oligonucleotide targets. Evaluation of selectivity for detection of single nucleotide polymorphisms (SNP) using such mixed films has now been completed. The selectivity was quantitatively determined by considering the sharpness of melt curves and melting temperature differences (DeltaT(m)) for fully complementary targets and SNPs. Stringency conditions were investigated, and it was determined that the selectivity was maximized when a moderate ionic strength was used (0.1-0.6 M). Increases of DeltaT(m) when using mixed films were up to 3-fold larger compared to surfaces containing only immobilized oligonucleotide probes. Concurrently, increases in sharpness of melt curves for 1 bpm targets were observed to be up to 2-fold greater for mixed films. The co-immobilization of PHEMA resulted in a more homogeneous distribution of oligonucleotide probes on surfaces. Lifetime measurements of fluorescence emission from immobilized oligonucleotide probes labeled with Cy3 dye indicated the difference in microenvironment of immobilized oligonucleotides in the presence of PHEMA.

Fluorescence anisotropy: from single molecules to live cells.

The polarization of light emitted by fluorescent probes is an easily accessible physical quantity that is related to a multitude of molecular parameters including conformation, orientation, size and the nanoscale environment conditions, such as dynamic viscosity and temperature. In analytical biochemistry and analytical chemistry applied to biological problems, fluorescence anisotropy is widely used for measuring the folding state of proteins and nucleic acids, and the affinity constant of ligands through titration experiments. The emphasis of this review is on new multi-parameter single-molecule detection schemes and their bioanalytical applications, and on the use of ensemble polarization assays to study binding and conformational dynamics of proteins and aptamers and for high-throughput discovery of small-molecule drugs.

Coordination complex SH2 domain proteomimetics: an alternative approach to disrupting oncogenic protein-protein interactions.

We report the first application of coordination complexes as functional proteomimetics of the Src homology 2 (SH2) phosphopeptide-binding domain. As a proof-of-concept, functionalized bis-dipicolylamine (BDPA) copper(ii) complexes are shown to disrupt oncogenic Stat3-Stat3 protein complexes and elicit promising anti-tumour activity.

Fluorescence depolarization studies of filamentous actin analyzed with a genetic algorithm.

A new method, in which a genetic algorithm was combined with Brownian dynamics and Monte Carlo simulations, was developed to analyze fluorescence depolarization data collected by the time-correlated single photon-counting technique. It was applied to studies of BODIPY-labeled filamentous actin (F-actin). The technique registered the local order and reorienting motions of the fluorophores, which were covalently coupled to cysteine 374 (C374) in actin and interacted by electronic energy migration within the actin polymers. Analyses of F-actin samples composed of different fractions of labeled actin molecules revealed the known helical organization of F-actin, demonstrating the usefulness of this technique for structure determination of complex protein polymers. The distance from the filament axis to the fluorophore was found to be considerably less than expected from the proposed position of C374 at a high filament radius. In addition, polymerization experiments with BODIPY-actin suggest a 25-fold more efficient signal for filament formation than pyrene-actin.

Self-aggregation--an intrinsic property of G(M1) in lipid bilayers.

We demonstrate that the ganglioside G(M1) in lipid bilayers of 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) exhibits a non-uniform lateral distribution, i.e., enriched regions of GM(1) molecules are formed, which is an argument in favour of self-aggregation of G(M1) being an intrinsic property of G(M1) ganglioside. This was concluded from energy transfer/migration studies of BODIPY-labelled gangliosides by means of time-resolved fluorescence lifetime and depolarization experiments. Three fluorophore-labelled gangliosides were synthesized to include either of two spectroscopically different BODIPY groups. These were specifically localized either in the polar headgroup region or in the non-polar region of the lipid bilayer. An eventual ganglioside-ganglioside affinity/aggregation induced by the BODIPY groups was experimentally excluded, which suggests their use in examining the influence of G(M1) in more complex systems.

Pyrromethene dyes (BODIPY) can form ground state homo and hetero dimers: photophysics and spectral properties.

Homo and hetero dimerisation of two spectroscopically different BODIPY chromophores was studied, namely, 4,4-difluoro-4-bora-3a,4a-diaza-s-indacene and its 5-styryl-derivative. These exhibit very similar absorption and fluorescence spectral shape, but are mutually shifted by ca. 70 nm. For this reason the former and the latter are referred to as the green and red BODIPY, which here are denoted gB and rB, respectively. Various spectroscopic properties of the rB in different common solvents were determined. The calculated and experimental fluorescence quantum yield is found to be close to 100%, the fluorescence relaxation has a single exponential decay with a lifetime of about 4.5 ns, and the Förster radius for donor-donor energy migration is 67+/-1A. The dimerisation in different solvents was examined by using custom synthesised; mono and bis BODIPY-labelled forms of 1,2-cis-diaminocyclohexane. It is shown that gB and rB can form ground state homo- as well as hetero dimers. The dimers are non-fluorescent, compatible with H-dimers and may act as excitation traps or as acceptors to the corresponding excited monomers.

On the quantitative treatment of donor-donor energy migration in regularly aggregated proteins.

An algorithm is presented that quantitatively accounts for donor-donor energy migration (DDEM) among fluorophore-labeled proteins forming regular aggregates. The DDEM algorithm is based on Monte Carlo and Brownian dynamics simulations and applies to calculation of fluorescence depolarisation data, such as the fluorescence anisotropy. Thereby local orientations, as well as reorienting motions of the fluorescent group are considered in the absence and presence of DDEM and among, in principle, infinitely many proteins as they form regular aggregates. Here we apply the algorithm for calculating and illustrating the DDEM and the time-resolved fluorescence anisotropy under static as well as dynamic conditions within helical, linear and circular aggregate structures. A principal approach of the DDEM algorithm for analysing protein aggregates is also outlined.