The Experimentalist's Guide to Machine Learning for Small Molecule Design.

Initially part of the field of artificial intelligence, machine learning (ML) has become a booming research area since branching out into its own field in the 1990s. After three decades of refinement, ML algorithms have accelerated scientific developments across a variety of research topics. The field of small molecule design is no exception, and an increasing number of researchers are applying ML techniques in their pursuit of discovering, generating, and optimizing small molecule compounds. The goal of this review is to provide simple, yet descriptive, explanations of some of the most commonly utilized ML algorithms in the field of small molecule design along with those that are highly applicable to an experimentally focused audience. The algorithms discussed here span across three ML paradigms: supervised learning, unsupervised learning, and ensemble methods. Examples from the published literature will be provided for each algorithm. Some common pitfalls of applying ML to biological and chemical data sets will also be explained, alongside a brief summary of a few more advanced paradigms, including reinforcement learning and semi-supervised learning.

Spin-Based Chiral Separations and the Importance of Molecule-Solvent Interactions.

This work uses magneto-electrochemical quartz crystal microbalance methods to study the enantiospecific adsorption of chiral molecules onto a ferromagnetic substrate. The effects of solution conditions, pH, and solvent isotope composition indicate that the kinetics of the enantiomeric adsorption depend strongly on the charge state and geometry of the adsorbate, whereas no thermodynamic contributions to enantiospecificity are found. Density functional theory calculations reveal that an interplay between the adsorbate and solvent molecules is important for defining the observed enantiospecific preference with an applied magnetic field; however, it remains unclear if intermolecular vibrational couplings contribute to the phenomenon.

Beyond Stereoisomeric Effects: Exploring the Importance of Intermolecular Electron Spin Interactions in Biorecognition.

This work shows that electron spin polarization and stereoisomeric effects make comparable contributions to the enantioselective binding of amino acids. Magneto-electrochemical quartz crystal microbalance methods are used to study the adsorption of chiral amino acids onto a monolayer film of chiral molecules that are spin polarized by an underlying ferromagnetic substrate. The direction of the electron spin polarization affects both the kinetics and thermodynamics of the enantiospecific adsorption of the amino acids. Comparison of these data with the circular dichroism (CD) spectra of the amino acid adsorbates shows that the CD spectrum of the interacting group provides a good figure-of-merit for predicting the contributions of electron spin to the intermolecular interaction. These findings demonstrate the importance of electron spin in enantioselective intermolecular interactions between chiral amino acids and represent a paradigm shift for how selectivity should be viewed in biorecognition.

Spin Selectivity Damage Dependence of Adsorption of dsDNA on Ferromagnets.

The adsorption of oxidatively damaged DNA onto ferromagnetic substrates was investigated. Both confocal fluorescence microscopy and quartz crystal microbalance methods show that the adsorption rate and the coverage depend on the magnetization direction of the substrate and the position of the damage site on the DNA relative to the substrate. SQUID magnetometry measurements show that the subsequent magnetic susceptibility of the DNA-coated ferromagnetic film depends on the direction of the magnetic field that was applied to the ferromagnetic film as the molecules were adsorbed. This study reveals that (i) the spin and charge polarization in DNA molecules is changed significantly by oxidative damage in the G bases and (ii) the rate of adsorption on a ferromagnet, as a function of the direction of the magnetic dipole of the surface, can be used as an assay to detect oxidative damage in the DNA.

Physics-model-based neural networks for inverse design of binary phase planar diffractive lenses.

The inverse design approach has enabled the customized design of photonic devices with engineered functionalities through adopting various optimization algorithms. However, conventional optimization algorithms for inverse design encounter difficulties in multi-constrained problems due to the substantial time consumed in the random searching process. Here, we report an efficient inverse design method, based on physics-model-based neural networks (PMNNs) and Rayleigh-Sommerfeld diffraction theory, for engineering the focusing behavior of binary phase planar diffractive lenses (BPPDLs). We adopt the proposed PMNN to design BPPDLs with designable functionalities, including realizing a single focal spot, multiple foci, and an optical needle with size approaching the diffraction limit. We show that the time for designing single device is dramatically reduced to several minutes. This study provides an efficient inverse method for designing photonic devices with customized functionalities, overcoming the challenges based on traditional data-driven deep learning.

Polymerized Hybrid Perovskites with Enhanced Stability, Flexibility, and Lattice Rigidity.

The intrinsic soft lattice nature of organometal halide perovskites (OHPs) makes them very tolerant to defects and ideal candidates for solution-processed optoelectronic devices. However, the soft lattice results in low stability towards external stresses such as heating and humidity, high density of phonons and strong electron-phonon coupling (EPC). Here, it is demonstrated that the OHPs with unsaturated 4-vinylbenzylammonium (VBA) as organoammonium cations can be polymerized without damaging the perovskite structure and its tolerance to defects. The polymerized perovskites show enhanced stability and flexibility compared to regular three-dimensional and two-dimensional (2D) perovskites. Furthermore, the polymerized 4-vinylbenzylammonium group improves perovskite lattice rigidity substantially, resulting in a reduced non-radiative recombination rate because of suppressed electron-phonon coupling, and enhanced carrier mobility because of suppressed phonon scattering. 2D polymerized perovskite light-emitting diodes (PeLEDs) with strong electroluminescence at room temperature, and quasi-2D PeLEDs with an external quantum efficiency (EQE) of 23.2% and enhanced operation stability are demonstrated. The work has opened a new way of enhancing the intrinsic stability and optoelectronic properties of OHPs.

Monitoring reactivation of latent HIV by label-free gradient light interference microscopy.

Human immunodeficiency virus (HIV) can infect cells and take a quiescent and nonexpressive state called latency. In this study, we report insights provided by label-free, gradient light interference microscopy (GLIM) about the changes in dry mass, diameter, and dry mass density associated with infected cells that occur upon reactivation. We discovered that the mean cell dry mass and mean diameter of latently infected cells treated with reactivating drug, TNF-α, are higher for latent cells that reactivate than those of the cells that did not reactivate. Cells with mean dry mass and diameter less than approximately 10 pg and 8 µm, respectively, remain exclusively in the latent state. Also, cells with mean dry mass greater than approximately 28-30 pg and mean diameter greater than 11-12 µm have a higher probability of reactivating. This study is significant as it presents a new label-free approach to quantify latent reactivation of a virus in single cells.

A transient heritable memory regulates HIV reactivation from latency.

Reactivation of human immunodeficiency virus 1 (HIV-1) from latently infected T cells is a critical barrier to cure patients. It remains unknown whether reactivation of individual latent cells occurs stochastically in response to latency reversal agents (LRAs) or is a deterministic outcome of an underlying cell state. To characterize these single-cell responses, we leverage the classical Luria-Delbrück fluctuation test where single cells are isolated from a clonal population and exposed to LRAs after colony expansion. Data show considerable colony-to-colony fluctuations with the fraction of reactivating cells following a skewed distribution. Modeling systematic measurements of fluctuations over time uncovers a transient heritable memory that regulates HIV-1 reactivation, where single cells are in an LRA-responsive state for a few weeks before switching back to an irresponsive state. These results have enormous implications for designing therapies to purge the latent reservoir and further utilize fluctuation-based assays to uncover hidden transient cellular states underlying phenotypic heterogeneity.

Screening for gene expression fluctuations reveals latency-promoting agents of HIV.

Upon treatment removal, spontaneous reactivation of latently infected T cells remains a major barrier toward curing HIV. Therapies that reactivate and clear the latent reservoir are only partially effective, while latency-promoting agents (LPAs) used to suppress reactivation and stabilize latency are understudied and lack diversity in their mechanisms of action. Here, we identify additional LPAs using a screen for gene-expression fluctuations (or "noise") that drive cell-fate specification and control HIV reactivation from latency. Single-cell protein dynamics of a minimal HIV gene circuit were monitored with time-lapse fluorescence microscopy. We screened 1,806 drugs, out of which 279 modulate noise magnitude or half autocorrelation time. Next, we tested the strongest noise modulators in a Jurkat T cell latency model and discovered three LPAs that would be overlooked by quantifying their mean expression levels alone. The LPAs reduced reactivation of latency in both Jurkat and primary cell models when challenged by synergistic and potent combinations of HIV activators. The two strongest LPAs, NSC 401005 and NSC 400938, are structurally and functionally related to inhibitors of thioredoxin reductase, a protein involved in maintaining redox balance in host cells. Experiments with multiple functional analogs revealed two additional LPAs, PX12 and tiopronin, and suggest a potential LPA family, within which some are commercially available and Food and Drug Administration-approved. The LPAs presented here may provide new strategies to complement antiretroviral treatments. Screening for gene expression noise holds the potential for drug discovery in other diseases.

High expression of TXNDC11 indicated unfavorable prognosis of glioma.

BACKGROUND: Thioredoxin domain containing 11 (TXNDC11) has been implicated in numerous cancers. Nevertheless, the function of TXNDC11 in glioma is not well described. This study aimed to assess clinical significance of TXNDC11 in glioma based on bioinformatics analysis and immunohistochemical (IHC) staining. METHODS: GEPIA2, The Cancer Genome Atlas (TCGA), and Gene Expression Omnibus (GEO) databases were employed to detect the levels of TXNDC11 transcript in glioma. Gene expression profiles and data from the methylation chip with clinical details from TCGA and Chinese Glioma Genome Atlas (CGGA) of glioma samples were examined. The methylation of TXNDC11 in glioma was evaluated by 450K methylation chip data analysis. The pathways involved in TXNDC11 expression were screened by gene set enrichment analysis (GSEA). The correlation between TXNDC11 and immune cells was analyzed. Protein level of TXNDC11 was detected by IHC staining in glioma specimens. RESULTS: TXNDC11 was highly expressed in glioma, and high TXNDC11 expression was associated with poor overall survival (OS) and worse clinical prognostic variables. The methylation of cg04399632 was statistically different between glioma samples and normal samples, and was negatively correlated with TXNDC11 expression in glioma patients. Survival analysis demonstrated a poorer prognosis in glioma patients with cg04399632 hypomethylation. TXNDC11-high phenotype was associated with certain immune-related pathways and other signaling pathways in glioma. The expression of TXNDC11 was correlated positively with M2 macrophage infiltration and negatively with M0 and M1 macrophage infiltration. IHC staining confirmed that TXNDC11 expression increased in higher-grade glioma. CONCLUSIONS: High expression of TXNDC11 may predict unfavorable prognosis of glioma patients.

Inhibition of the mechanistic target of rapamycin induces cell survival via MAPK in tuberous sclerosis complex.

BACKGROUND: Tuberous sclerosis complex (TSC) is a genetic disorder that cause tumors to form in many organs. These lesions may lead to epilepsy, autism, developmental delay, renal, and pulmonary failure. Loss of function mutations in TSC1 and TSC2 genes by aberrant activation of the mechanistic target of rapamycin (mTORC1) signaling pathway are the known causes of TSC. Therefore, targeting mTORC1 becomes a most available therapeutic strategy for TSC. Although mTORC1 inhibitor rapamycin and Rapalogs have demonstrated exciting results in the recent clinical trials, however, tumors rebound and upon the discontinuation of the mTORC1 inhibition. Thus, understanding the underlying molecular mechanisms responsible for rapamycin-induced cell survival becomes an urgent need. Identification of additional molecular targets and development more effective remission-inducing therapeutic strategies are necessary for TSC patients. RESULTS: We have discovered an Mitogen-activated protein kinase (MAPK)-evoked positive feedback loop that dampens the efficacy of mTORC1 inhibition. Mechanistically, mTORC1 inhibition increased MEK1-dependent activation of MAPK in TSC-deficient cells. Pharmacological inhibition of MAPK abrogated this feedback loop activation. Importantly, the combinatorial inhibition of mTORC1 and MAPK induces the death of TSC2-deficient cells. CONCLUSIONS: Our results provide a rationale for dual targeting of mTORC1 and MAPK pathways in TSC and other mTORC1 hyperactive neoplasm.

Estrogen activates pyruvate kinase M2 and increases the growth of TSC2-deficient cells.

Lymphangioleiomyomatosis (LAM) is a devastating lung disease caused by inactivating gene mutations in either TSC1 or TSC2 that result in hyperactivation of the mechanistic target of rapamycin complex 1 (mTORC1). As LAM occurs predominantly in women during their reproductive age and is exacerbated by pregnancy, the female hormonal environment, and in particular estrogen, is implicated in LAM pathogenesis and progression. However, detailed underlying molecular mechanisms are not well understood. In this study, utilizing human pulmonary LAM specimens and cell culture models of TSC2-deficient LAM patient-derived and rat uterine leiomyoma-derived cells, we tested the hypothesis that estrogen promotes the growth of mTORC1-hyperactive cells through pyruvate kinase M2 (PKM2). Estrogen increased the phosphorylation of PKM2 at Ser37 and induced the nuclear translocation of phospho-PKM2. The estrogen receptor antagonist Faslodex reversed these effects. Restoration of TSC2 inhibited the phosphorylation of PKM2 in an mTORC1 inhibitor-insensitive manner. Finally, accumulation of phosphorylated PKM2 was evident in pulmonary nodule from LAM patients. Together, our data suggest that female predominance of LAM might be at least in part attributed to estrogen stimulation of PKM2-mediated cellular metabolic alterations. Targeting metabolic regulators of PKM2 might have therapeutic benefits for women with LAM and other female-specific neoplasms.

Insights into the precursor effect on the surface structure of γ-Al2O3 and NO + CO catalytic performance of CO-pretreated CuO/MnOx/γ-Al2O3 catalysts.

NO reduction by CO was investigated over CO-pretreated CuO/MnOx/γ-Al2O3 catalysts with different metal precursors (nitrate and acetate). It was found that the catalyst prepared from acetate salts (Cu/Mn/Al-A) exhibited significantly higher activity than counterpart catalyst from nitrate precursors (Cu/Mn/Al-N). XRD, XPS and in situ DRIFT were carried out to approach the nature for the different catalytic performance. For both catalysts, copper mainly existed as CuO, but the status of manganese oxide was markedly different. Mn(IV) was predominant in Cu/Mn/Al-N and Mn(III) was enriched in Cu/Mn/Al-A. As a result, different dispersion behaviors of manganese oxide on γ-Al2O3 were displayed, which induced inconsistent Cu-Mn contact. The catalyst obtained from acetate precursor exhibited enriched Cu-Mn contact and thus more Cu+-□-Mn3+/2+ entities would be produced after CO pretreatment, leading to promoted NO dissociation and favorable performance in NO reduction by CO. The present study sheds light on the effective tuning of Cu-O-Mn interfacial sites in CuO/MnOx/γ-Al2O3 via modulating the dispersion behaviors of surface components.

Rapalog resistance is associated with mesenchymal-type changes in Tsc2-null cells.

Tuberous Sclerosis Complex (TSC) and Lymphangioleiomyomatosis (LAM) are caused by inactivating mutations in TSC1 or TSC2, leading to mTORC1 hyperactivation. The mTORC1 inhibitors rapamycin and analogs (rapalogs) are approved for treating of TSC and LAM. Due to their cytostatic and not cytocidal action, discontinuation of treatment leads to tumor regrowth and decline in pulmonary function. Therefore, life-long rapalog treatment is proposed for the control of TSC and LAM lesions, which increases the chances for the development of acquired drug resistance. Understanding the signaling perturbations leading to rapalog resistance is critical for the development of better therapeutic strategies. We developed the first Tsc2-null rapamycin-resistant cell line, ELT3-245, which is highly tumorigenic in mice, and refractory to rapamycin treatment. In vitro ELT3-245 cells exhibit enhanced anchorage-independent cell survival, resistance to anoikis, and loss of epithelial markers. A key alteration in ELT3-245 is increased ß-catenin signaling. We propose that a subset of cells in TSC and LAM lesions have additional signaling aberrations, thus possess the potential to become resistant to rapalogs. Alternatively, when challenged with rapalogs TSC-null cells are reprogrammed to express mesenchymal-like markers. These signaling changes could be further exploited to induce clinically-relevant long-term remissions.

Cell Size-Based Decision-Making of a Viral Gene Circuit.

Latently infected T cells able to reinitiate viral propagation throughout the body remain a major barrier to curing HIV. Distinguishing between latently infected cells and uninfected cells will advance efforts for viral eradication. HIV decision-making between latency and active replication is stochastic, and drug cocktails that increase bursts of viral gene expression enhance reactivation from latency. Here, we show that a larger host-cell size provides a natural cellular mechanism for enhancing burst size of viral expression and is necessary to destabilize the latent state and bias viral decision-making. Latently infected Jurkat and primary CD4+ T cells reactivate exclusively in larger activated cells, while smaller cells remain silent. In addition, reactivation is cell-cycle dependent and can be modulated with cell-cycle-arresting compounds. Cell size and cell-cycle dependent decision-making of viral circuits may guide stochastic design strategies and applications in synthetic biology and may provide important determinants to advance diagnostics and therapies.

Synthesis of CrO x /C catalysts for low temperature NH3-SCR with enhanced regeneration ability in the presence of SO2.

Chromium oxide nano-particles with an average diameter of 3 nm covered by amorphous carbon (CrO x /C) were successfully synthesized. The synthesized CrO x /C materials were used for the selective catalytic reduction of NO x by NH3 (NH3-SCR), which shows superb NH3-SCR activity and in particular, satisfactory regeneration ability in the presence of SO2 compared with Mn-based catalysts. The as-prepared catalysts were characterized by XRD, HRTEM, Raman, FTIR, BET, TPD, TPR, XPS and in situ FTIR techniques. The results indicated presence of certain amounts of unstable lattice oxygen exposed on the surface of CrO x nano-particles with an average size of 3 nm in the CrO x /C samples, which led to NO being conveniently oxidized to NO2. The formed NO2 participated in NH3-SCR activity, reacting with catalysts via a "fast NH3-SCR" pathway, which enhanced th NH3-SCR performance of the CrO x /C catalysts. Furthermore, the stable lattice of the CrO x species made the catalyst immune to the sulfation process, which was inferred to be the cause of its superior regeneration ability in the presence of SO2. This study provides a simple way to synthesize stable CrO x nano-particles with active oxygen, and sheds light on designing NH3-SCR catalysts with highly efficient low temperature activity, SO2 tolerance, and regeneration ability.

Fine-tuning of noise in gene expression with nucleosome remodeling.

Engineering stochastic fluctuations of gene expression (or "noise") is integral to precisely bias cellular-fate decisions and statistical phenotypes in both single-cell and multi-cellular systems. Epigenetic regulation has been shown to constitute a large source of noise, and thus, engineering stochasticity is deeply intertwined with epigenetics. Here, utilizing chromatin remodeling, we report that Caffeic acid phenethyl ester (CA) and Pyrimethamine (PYR), two inhibitors of BAF250a, a subunit of the Brahma-associated factor (BAF) nucleosome remodeling complex, enable differential and tunable control of noise in transcription and translation from the human immunodeficiency virus long terminal repeat promoter in a dose and time-dependent manner. CA conserves noise levels while increasing mean abundance, resulting in direct tuning of the transcriptional burst size, while PYR strictly increases transcriptional initiation frequency while conserving a constant transcriptional burst size. Time-dependent treatment with CA reveals non-continuous tuning with noise oscillating at a constant mean abundance at early time points and the burst size increasing for treatments after 5 h. Treatments combining CA and Protein Kinase C agonists result in an even larger increase of abundance while conserving noise levels with a highly non-linear increase in variance of up to 63× untreated controls. Finally, drug combinations provide non-antagonistic combinatorial tuning of gene expression noise and map a noise phase space for future applications with viral and synthetic gene vectors. Active remodeling of nucleosomes and BAF-mediated control of gene expression noise expand a toolbox for the future design and engineering of stochasticity in living systems.

Effect of precursors on the structure and activity of CuO-CoOx/γ-Al2O3 catalysts for NO reduction by CO.

Catalytic reduction of NO by CO was studied over a series of CuO-CoOx/γ-Al2O3 catalysts prepared by co-impregnation with different copper and cobalt precursors (acetate and nitrate) to evaluate the structure-activity relationship. The obtained samples were characterized in detail by means of XRD, LRS, XPS, H2-TPR and in situ FT-IR technologies. Results indicate that copper oxide is agglomerated while cobalt oxide is dispersed on γ-Al2O3 for the catalyst prepared from copper acetate and cobalt acetate precursors (CuACoA); CuxCo3-xO4 spinel is formed and agglomerated on the catalyst prepared from copper nitrate and cobalt nitrate precursors (CuNCoN); while both copper oxide and cobalt oxide could be homogeneously dispersed for the catalyst prepared from copper nitrate and cobalt acetate precursors (CuNCoA), which exhibits the best activity for NO reduction by CO. Probably the synergistic effect between dispersed copper oxide and cobalt oxide is propitious to the oxygen transfer, which could be the reason for its high activities. Finally, a possible reaction mechanism was tentatively proposed to explore the different catalytic performances in NO reduction by CO model reaction.

A novel sandwiched electrochemiluminescence immunosensor for the detection of carcinoembryonic antigen based on carbon quantum dots and signal amplification.

In this study, a novel sandwiched electrochemiluminescence (ECL) immunosensor for the detection of carcinoembryonic antigen (CEA) was developed. The nanocomposite of polydopamine and Ag nanoparticles (PDA-AgNPs) was prepared by the redox reaction between Ag+ and dopamine. This nanocomposite not only provided an effective matrix for the immobilization of primary antibody (Ab1) but also enhanced the conductivity of the electrode. Carbon quantum dots (CQDs) were immobilized on the poly(ethylenimine) functionalized graphene oxide (PEI-GO) through amido-bond. Then Au nanoparticles were decorated on the CQDs modified PEI-GO matrix, and the resulted complex AuNPs/CQDs-PEI-GO was introduced to link secondary antibody (Ab2). The CQDs can be connected to the electrode surface through the combination of CEA with Ab1 and Ab2, and then the amplified electrochemiluminescence signal of CQDs was obtained with the synergistic effect of AgNPs, polydopamine, AuNPs and PEI-GO. Under the optimal conditions, the ECL intensity was proportional to the logarithm value of CEA concentration in the linear range from 5pgmL-1 to 500ngmL-1 with a detection limit of 1.67pgmL-1 for CEA detection. The immunosensor was applied for the CEA detection in real samples with satisfactory results. The proposed ECL immunosensor showed good performance with high sensitivity, specificity, reproducibility, stability and will be potential in clinical detection.

Urine AQP5 is a potential novel biomarker of diabetic nephropathy.

AIMS: To investigate if urinary AQP5 serves as a new potential biomarker of diabetic nephropathy. METHODS: Using an AQP5-specific enzyme-linked immunosorbent assay, we measured serum and urine AQP5 first in a cohort consisting of normal controls (n=26) and patients with diabetes mellitus (n=25) or diabetic nephropathy (n=33) and then in a validation cohort possessing normal controls (n=10), patients with diabetes mellitus (n=10) or diabetic nephropathy (n=14), and patients with chronic kidney disease of unknown etiology (n=10). We used various statistical methods including Pearson's correlation coefficient, ANOVA with Holm-Sidak test, Receiver Operator Curve, and multiple logistic regression to analyze the data. RESULTS: Urine AQP5/creatinine 1) is significantly higher in diabetic nephropathy than in other two groups, and in diabetic nephropathy stage V than in stage III; 2) correlates with serum creatinine, urine albumin, and multiple other known risk factors of the disease; and 3) improves the clinical models in distinguishing diabetic nephropathy from normal controls and diabetic mellitus. CONCLUSION: Our data suggest that urine AQP5/creatinine may possess diagnostic and prognostic values as a biomarker of diabetic nephropathy.