Class-incremental learning with Balanced Embedding Discrimination Maximization.

Class incremental learning is committed to solving representation learning and classification assignments while avoiding catastrophic forgetting in scenarios where categories are increasing. In this work, a unified method named Balanced Embedding Discrimination Maximization (BEDM) is developed to make the intermediate embedding more distinctive. Specifically, we utilize an orthogonality constraint based on doubly-blocked Toeplitz matrix to minimize the correlation of convolution kernels, and an algorithm for similarity visualization is introduced. Furthermore, uneven samples and distribution shift among old and new tasks eventuate strongly biased classifiers. To mitigate the imbalance, we propose an adaptive balance weighting in softmax to compensate insufficient categories dynamically. In addition, hybrid embedding learning is introduced to preserve knowledge from old models, which involves less hyper-parameters than conventional knowledge distillation. Our proposed method outperforms the existing approaches on three mainstream benchmark datasets. Moreover, we technically visualize that our method can produce a more uniform similarity histogram and more stable spectrum. Grad-CAM and t-SNE visualizations further confirm its effectiveness. Code is available at https://github.com/wqzh/BEDM.

Precision Engineering of the Co-immobilization of Enzymes for Cascade Biocatalysis.

The design and orderly layered co-immobilization of multiple enzymes on resin particles remain challenging. In this study, the SpyTag/SpyCatcher binding pair was fused to the N-terminus of an alcohol dehydrogenase (ADH) and an aldo-keto reductase (AKR), respectively. A non-canonical amino acid (ncAA), p-azido-L-phenylalanine (p-AzF), as the anchor for covalent bonding enzymes, was genetically inserted into preselected sites in the AKR and ADH. Employing the two bioorthogonal counterparts of SpyTag/SpyCatcher and azide-alkyne cycloaddition for the immobilization of AKR and ADH enabled sequential dual-enzyme coating on porous microspheres. The ordered dual-enzyme reactor was subsequently used to synthesize (S)-1-(2-chlorophenyl)ethanol asymmetrically from the corresponding prochiral ketone, enabling the in situ regeneration of NADPH. The reactor exhibited a high catalytic conversion of 74 % and good reproducibility, retaining 80 % of its initial activity after six cycles. The product had 99.9 % ee, which that was maintained in each cycle. Additionally, the double-layer immobilization method significantly increased the enzyme loading capacity, which was approximately 1.7 times greater than that of traditional single-layer immobilization. More importantly, it simultaneously enabled both the purification and immobilization of multiple enzymes on carriers, thus providing a convenient approach to facilitate cascade biocatalysis.

Reflections on a Copenhagen-Minneapolis Axis in Bioorganic Chemistry.

The international peptide community rejoiced when one of its most distinguished members, Morten Meldal of Denmark, shared the 2022 Nobel Prize in Chemistry. In fact, the regiospecific solid-phase "copper(I)-catalyzed 1,3-dipolar cycloaddition of terminal alkynes to azides" (CuACC) reaction-that formed the specific basis for Meldal's recognition-was reported first at the 17th American Peptide Symposium held in San Diego in June 2001. The present perspective outlines intertwining conceptual and experimental threads pursued concurrently in Copenhagen and Minneapolis, sometimes by the same individuals, that provided context for Meldal's breakthrough discovery. Major topics covered include orthogonality in chemistry; the dithiasuccinoyl (Dts) protecting group for amino groups in α-amino acids, carbohydrates, and monomers for peptide nucleic acids (PNA); and poly(ethylene glycol) (PEG)-based solid supports such as PEG-PS, PEGA, and CLEAR [and variations inspired by them] for solid-phase peptide synthesis (SPPS), solid-phase organic synthesis (SPOS), and combinatorial chemistry that can support biological assays in aqueous media.

Triplet-Triplet Energy Transfer: A Simple Strategy for an Efficient Visible Light-Induced Photoclick Reaction.

Photoclick reactions combine the advantages offered by light-driven processes and classical click chemistry and have found applications ranging from surface functionalization, polymer conjugation, photo-crosslinking, and protein labeling. Despite these advances, the dependency of most of the photoclick reactions on UV light poses a severe obstacle for their general implementation, as this light can be absorbed by other molecules in the system resulting in their degradation or unwanted reactivity. However, the development of a simple and efficient system to achieve bathochromically shifted photoclick transformations remains challenging. Here, we introduce triplet-triplet energy transfer as a fast and selective way to enable visible light-induced photoclick reactions. Specifically, we show that 9,10-phenanthrenequinones (PQs) can efficiently react with electron-rich alkenes (ERAs) in the presence of a catalytic amount (as little as 5 mol %) of photosensitizers. The photocycloaddition reaction can be achieved under green (530 nm) or orange (590 nm) light irradiation, representing a bathochromic shift of over 100 nm as compared to the classical PQ-ERAs system. Furthermore, by combining appropriate reactants, we establish an orthogonal, blue and green light-induced photoclick reaction system in which the product distribution can be precisely controlled by the choice of the color of light.

Revisiting column selectivity choices in ultra-high performance liquid chromatography-Using multidimensional analytical Design Spaces to identify column equivalency.

Current guides and column selection system (CSS) platforms can provide some helpful insights with regard to the selection of alternative phases. Their practical reliability however, can also turn out to be questionable, especially considering the lack of detailed specifics, such as a clear definition of points of equivalence-appropriate running conditions under which the given analytical mixture can be satisfactorily resolved on various stationary phases. In this context, the use of multivariate modeling tools can be highly beneficial. These tools, when applied systematically, are ideal for uniquely characterizing complex LC-separation systems, a fact supported by numerous peer-reviewed papers. Revisiting our earlier work [1] and the applied systematic workflow [2], we used a Design Space modeling software (DryLab), with the main focus on building and comparing 3-dimensional separation models of amlodipine and its related impurities to identify shared method conditions under which columns are conveniently interchangeable. Our study comprised 5, C18-modified ultra-high performance liquid chromatography (UHPLC) columns in total, in some cases with surprising results. We identified several equivalences between the Design Spaces (DSs) of markedly different columns. Conversely, there were cases where, despite the predicted similarities in column data, the modeled DSs demonstrated clear differences between the selected stationary phases.

Orthogonal sp3-Ge/B Bimetallic Modules: Enantioselective Construction and Enantiospecific Cross-Coupling.

In this study, a series of enantioenriched sp3-Ge/B bimetallic modules were successfully synthesized via an enantioselective copper-catalyzed hydroboration of carbagermatrane (Ge)-containing alkenes. Orthogonal cross-coupling selectivity under different Pd-catalyzed conditions was achieved in an enantiospecific manner. Notably, the chiral secondary Ge exhibited a remarkable transmetallation ability prior to primary or secondary Bpin. The effectiveness of this Ge/B bimetallic strategy was further demonstrated through the development of new functional small molecules with Aggregation-Induced Emission (AIE) and Circularly Polarized Luminescence (CPL) performance. This represents the first successful example of synthesis of enantioenriched alkylgermanium reagents that permit enantiospecific cross-coupling reactions.

Topological Persistence Guided Knowledge Distillation for Wearable Sensor Data.

Deep learning methods have achieved a lot of success in various applications involving converting wearable sensor data to actionable health insights. A common application areas is activity recognition, where deep-learning methods still suffer from limitations such as sensitivity to signal quality, sensor characteristic variations, and variability between subjects. To mitigate these issues, robust features obtained by topological data analysis (TDA) have been suggested as a potential solution. However, there are two significant obstacles to using topological features in deep learning: (1) large computational load to extract topological features using TDA, and (2) different signal representations obtained from deep learning and TDA which makes fusion difficult. In this paper, to enable integration of the strengths of topological methods in deep-learning for time-series data, we propose to use two teacher networks - one trained on the raw time-series data, and another trained on persistence images generated by TDA methods. These two teachers are jointly used to distill a single student model, which utilizes only the raw time-series data at test-time. This approach addresses both issues. The use of KD with multiple teachers utilizes complementary information, and results in a compact model with strong supervisory features and an integrated richer representation. To assimilate desirable information from different modalities, we design new constraints, including orthogonality imposed on feature correlation maps for improving feature expressiveness and allowing the student to easily learn from the teacher. Also, we apply an annealing strategy in KD for fast saturation and better accommodation from different features, while the knowledge gap between the teachers and student is reduced. Finally, a robust student model is distilled, which can at test-time uses only the time-series data as an input, while implicitly preserving topological features. The experimental results demonstrate the effectiveness of the proposed method on wearable sensor data. The proposed method shows 71.74% in classification accuracy on GENEActiv with WRN16-1 (1D CNNs) student, which outperforms baselines and takes much less processing time (less than 17 sec) than teachers on 6k testing samples.

Recent Developments with Organogermanes: their Preparation and Application in Synthesis and Catalysis.

Within the sphere of traditional Pd0 /PdII cross coupling reactions, organogermanes have been historically outperformed both in terms of scope and reactivity by more conventional transmetalating reagents. Subsequently, this class of compounds has been largely underutilized as a coupling partner in bond-forming strategies. Most recent studies, however, have shown that alternative modes of activation of these notoriously robust building blocks transform organogermanes into the most reactive site of the molecule-capable of outcompeting other functional groups (such as boronic acids, esters and silanes) for both C-C and C-heteroatom bond formation. As a result, over the past few years, the literature has increasingly featured methodologies that explore the potential of organogermanes as chemoselective and orthogonal coupling partners. Herein we highlight some of these recent advances in the field of organogermane chemistry both with respect to their synthesis and applications in synthetic and catalytic transformations.

Green Solvent Selection for All Solution-Processed Inverted Quantum Dot Light Emitting Diode.

Quantum-dot light-emitting diodes (QD-LEDs) have gained attention as potential display technologies. However, the solvents used to dissolve a polymeric hole transport layer (HTL) are hazardous to both humans and the environment. Additionally, intermixing the HTL and QD layers presents a significant challenge when fabricating inverted QD-LEDs. Here, a green solvent selection procedure to achieve good device performance and environmental safety in QD-LEDs is established. This procedure utilizes Hansen solubility parameters and surface roughness to identify a set of solvents that do not lower the device performance by avoiding interlayer mixing or a rough interface. The CHEM21 solvent selection guide is used to screen for environmentally hazardous solvents. Finally, cyclopentanone (CPO) is selected as the optimal HTL solvent from among 16 candidates. Using CPO improves the maximum luminescence by ≈1.6 times and the maximum current efficiency by ≈12.6 times, compared to that of conventional devices using hazardous chlorobenzene. Solvent selection is critical for the fabrication of green and high-performance inverted QD-LEDs, particularly for large display panels that require n-type oxide thin-film transistors.

[Off-line comprehensive two-dimensional countercurrent chromatography-liquid chromatography separation of Curcuma volatile oil].

The chemical constituents of volatile oils used in traditional Chinese medicine are highly complex. Thus, achieving the complete separation of volatile oils by one-dimensional chromatography is difficult owing to the low peak capacity of the technique. Although comprehensive two-dimensional gas chromatography provides an efficient means for separating volatile oils, it cannot be used to screen bioactive components because of its limitations. Therefore, developing a new method to separate volatile oils based on liquid chromatography is of great significance in efforts to obtain new approaches to screen bioactive components in volatile oil. The objectives of the present study are to establish an efficient method for separating the chemical constituents of Curcuma volatile oil using off-line comprehensive two-dimensional countercurrent chromatography-liquid chromatography (CCC-LC) and to investigate the two-dimensional peak capacity, orthogonality, and spatial coverage of this method. Both CCC and LC conditions were optimized. A biphasic n-hexane-methanol-water solvent system was selected via the colorimetric method, and the lower phase was used as the mobile phase in gradient-elution mode: 0-55 min, n-hexane-methanol-water (5∶2∶3 v/v/v); 55-170 min, n-hexane-methanol-water (5∶3∶2, v/v/v); 170-290 min, n-hexane-methanol-water (5∶4∶1, v/v/v). After gradient elution, elution-extrusion elution mode was applied within 290-375 min. Good resolution was achieved by the CCC separation process. The HPLC separation process was carried out with gradient elution using a mobile phase composed of acetonitrile (A)-water (B): 0-10 min, 50%A-65%A; 10-14 min, 65%A; 14-21 min, 65%A-85%A; 21-25 min, 85%A-95%A; 25-30 min, 95%A-55%A; 30-40 min, 55%A. Curcuma volatile oil was separated under the above optimized two-dimensional separation conditions, and the data obtained were drawn into a two-dimensional contour map using Matlab software. The calculated total peak capacity exceeded 954, which was 10 times more than that of one-dimensional chromatography. High orthogonality (r=0.17) and spatial coverage factor (68.1%) were also obtained. Our research provides a new methodology for separating volatile oils used in traditional Chinese medicine as well as an approach for evaluating the quality of traditional Chinese medicinal herbs using two-dimensional chromatographic fingerprints.

Orthogonalized Kernel Debiased Machine Learning for Multimodal Data Analysis.

Multimodal imaging has transformed neuroscience research. While it presents unprecedented opportunities, it also imposes serious challenges. Particularly, it is difficult to combine the merits of the interpretability attributed to a simple association model with the flexibility achieved by a highly adaptive nonlinear model. In this article, we propose an orthogonalized kernel debiased machine learning approach, which is built upon the Neyman orthogonality and a form of decomposition orthogonality, for multimodal data analysis. We target the setting that naturally arises in almost all multimodal studies, where there is a primary modality of interest, plus additional auxiliary modalities. We establish the root-N-consistency and asymptotic normality of the estimated primary parameter, the semi-parametric estimation efficiency, and the asymptotic validity of the confidence band of the predicted primary modality effect. Our proposal enjoys, to a good extent, both model interpretability and model flexibility. It is also considerably different from the existing statistical methods for multimodal data integration, as well as the orthogonality-based methods for high-dimensional inferences. We demonstrate the efficacy of our method through both simulations and an application to a multimodal neuroimaging study of Alzheimer's disease.

Retention correlation and orthogonality between reversed phase countercurrent chromatography and liquid chromatography based on solvent strength.

Correlation of elution performance between reversed phase countercurrent chromatography and liquid chromatography was investigated using five selected natural components. Theoretical guidance for orthogonality of two-dimensional countercurrent chromatography and liquid chromatography was proposed. The difference in retention behavior between countercurrent chromatography and liquid chromatography was studied when the mobile phase was composed of methanol and water by measuring the partition behavior of five selected compounds in two typical biphasic solvent systems composed of n-hexane-ethyl acetate-methanol-water and chloroform-methanol-water. An orthogonal diagram between countercurrent chromatography and liquid chromatography was obtained by normalized treatment of the measured partition coefficients and capacity factors. The experimental results showed that each biphasic solvent system used for countercurrent chromatography had a high orthogonality with liquid chromatography when a specific volume ratio was used.

Evaluating orthogonality between ion-pair reversed phase, anion exchange, and hydrophilic interaction liquid chromatography for the separation of synthetic oligonucleotides.

The orthogonality of separation between ion-pair reversed phase (IP-RP), anion exchange (AEX), and hydrophilic interaction liquid chromatography (HILIC) was evaluated for oligonucleotides. A polythymidine standard ladder was first used to evaluate the three methods and showed zero orthogonality, where retention and selectivity were based on oligonucleotide charge/size under all three conditions. Next, a model 23-mer synthetic oligonucleotide containing 4 phosphorothioate bonds with 2' fluoro and 2'-O-methyl ribose modifications typical of small interfering RNA was used for evaluating orthogonality. The resolution and orthogonality were evaluated between the three modes of chromatography in terms of selectivity differences for nine common impurities, including truncations (n-1, n-2), addition (n + 1), oxidation, and de-fluorination. We first evaluated different ion-pairing reagents that provided the best separation of the key impurities while suppressing diastereomer separation due to phosphorothioate linkages. Although different ion-pairing reagents affected resolution, very little orthogonality was observed. We then compared the retention times between IP-RP, HILIC, and AEX for each impurity of the model oligonucleotide and observed various selectivity changes. The results suggest that coupling HILIC with either AEX or IP-RP provide the highest degree of orthogonality due to the differences in retention for hydrophilic nucleobases and modifications under HILIC conditions. IP-RP provided the highest overall resolution for the impurity mixture, whereas more co-elution was observed with HILIC and AEX. The unique selectivity patterns offered by HILIC provides an interesting alternative to IP-RP or AEX, in addition to the potential for coupling with multidimensional separations. Future work should explore orthogonality for oligonucleotides with subtle sequence differences such as nucleobase modifications and base flip isomers, longer strands such as guide RNA and messenger RNA, and other biotherapeutic modalities such as peptides, antibodies, and antibody-drug-conjugates.

On Graphical Models and Convex Geometry.

A mixture-model of beta distributions framework is introduced to identify significant correlations among P features when P is large. The method relies on theorems in convex geometry, which are used to show how to control the error rate of edge detection in graphical models. The proposed 'betaMix' method does not require any assumptions about the network structure, nor does it assume that the network is sparse. The results hold for a wide class of data-generating distributions that include light-tailed and heavy-tailed spherically symmetric distributions. The results are robust for sufficiently large sample sizes and hold for non-elliptically-symmetric distributions.

Leaderless Bicistronic Design for Precise and Reliable Control of Gene Expression in Corynebacterium Glutamicum.

In synthetic biology, the precise control of gene expression is challenging due to the limited orthogonality of expression elements. Here, to address this issue and improve the reusability of genetic elements, we developed a bicistronic expression cassette in Corynebacterium glutamicum based on a leaderless promoter lacking a 5'UTR. The created leaderless bicistronic design (BCD) significantly improved the orthogonality of expression elements across different genes of interest. We also explored the importance of the fore-cistron and SD motif in maintaining the strength of leaderless BCDs. Additionally, we established a library containing 55,901 fore-cistrons and demonstrated that the regulatory range of gene expression in leaderless BCDs can be broader by modifying the fore-cistron sequence. This study provides a novel synthetic biology tool based on leaderless BCD for fine-tuning gene expression in C. glutamicum using fore-cistrons. Moreover, the strategy developed here can also be applied to improve the performance of other leaderless promoters in other bacteria.

Reversed HILIC Gradient: A Powerful Strategy for On-Line Comprehensive 2D-LC.

The aim of the present work is to evaluate the possibilities and limitations of reversed hydrophilic interaction chromatography (revHILIC) mode in liquid chromatography (LC). This chromatographic mode consists of combining a highly polar stationary phase (bare silica) with a gradient varying from very low (1-5%) to high (40%) acetonitrile content (reversed gradient compared to HILIC). The retention behavior of revHILIC was first compared with that of reversed-phase LC (RPLC) and HILIC using representative mixtures of peptides and pharmaceutical compounds. It appears that the achievable selectivity can be ranked in the order RPLC > revHILIC > HILIC with the two different samples. Next, two-dimensional liquid chromatography (2D-LC) conditions were evaluated by combining RPLC, revHILIC, or HILIC with RPLC in an on-line comprehensive (LC × LC) mode. evHILIC × RPLC not only showed impressive performance in terms of peak capacity and sensitivity, but also provided complementary selectivity compared to RPLC × RPLC and HILIC × RPLC. Indeed, both the elution order and the retention time range differ significantly between the three techniques. In conclusion, there is no doubt that revHILIC should be considered as a viable option for 2D-LC analysis of small molecules and also peptides.

The Expanded Central Dogma: Genome Resynthesis, Orthogonal Biosystems, Synthetic Genetics.

Synthetic biology seeks to probe fundamental aspects of biological form and function by construction [i.e., (re)synthesis] rather than deconstruction (analysis). In this sense, biological sciences now follow the lead given by the chemical sciences. Synthesis can complement analytic studies but also allows novel approaches to answering fundamental biological questions and opens up vast opportunities for the exploitation of biological processes to provide solutions for global problems. In this review, we explore aspects of this synthesis paradigm as applied to the chemistry and function of nucleic acids in biological systems and beyond, specifically, in genome resynthesis, synthetic genetics (i.e., the expansion of the genetic alphabet, of the genetic code, and of the chemical make-up of genetic systems), and the elaboration of orthogonal biosystems and components.

Evolutionary paths that link orthogonal pairs of binding proteins.

Some protein binding pairs exhibit extreme specificities that functionally insulate them from homologs. Such pairs evolve mostly by accumulating single-point mutations, and mutants are selected if their affinity exceeds the threshold required for function1-4. Thus, homologous and high-specificity binding pairs bring to light an evolutionary conundrum: how does a new specificity evolve while maintaining the required affinity in each intermediate5,6? Until now, a fully functional single-mutation path that connects two orthogonal pairs has only been described where the pairs were mutationally close thus enabling experimental enumeration of all intermediates2. We present an atomistic and graph-theoretical framework for discovering low molecular strain single-mutation paths that connect two extant pairs, enabling enumeration beyond experimental capability. We apply it to two orthogonal bacterial colicin endonuclease-immunity pairs separated by 17 interface mutations7. We were not able to find a strain-free and functional path in the sequence space defined by the two extant pairs. But including mutations that bridge amino acids that cannot be exchanged through single-nucleotide mutations led us to a strain-free 19-mutation trajectory that is completely viable in vivo. Our experiments show that the specificity switch is remarkably abrupt, resulting from only one radical mutation on each partner. Furthermore, each of the critical specificity-switch mutations increases fitness, demonstrating that functional divergence could be driven by positive Darwinian selection. These results reveal how even radical functional changes in an epistatic fitness landscape may evolve.

Assessing the Orthogonality of Phage-Encoded RNA Polymerases for Tailored Synthetic Biology Applications in Pseudomonas Species.

The phage T7 RNA polymerase (RNAP) and lysozyme form the basis of the widely used pET expression system for recombinant expression in the biotechnology field and as a tool in microbial synthetic biology. Attempts to transfer this genetic circuitry from Escherichia coli to non-model bacterial organisms with high potential have been restricted by the cytotoxicity of the T7 RNAP in the receiving hosts. We here explore the diversity of T7-like RNAPs mined directly from Pseudomonas phages for implementation in Pseudomonas species, thus relying on the co-evolution and natural adaptation of the system towards its host. By screening and characterizing different viral transcription machinery using a vector-based system in P. putida., we identified a set of four non-toxic phage RNAPs from phages phi15, PPPL-1, Pf-10, and 67PfluR64PP, showing a broad activity range and orthogonality to each other and the T7 RNAP. In addition, we confirmed the transcription start sites of their predicted promoters and improved the stringency of the phage RNAP expression systems by introducing and optimizing phage lysozymes for RNAP inhibition. This set of viral RNAPs expands the adaption of T7-inspired circuitry towards Pseudomonas species and highlights the potential of mining tailored genetic parts and tools from phages for their non-model host.

Molecular Engineering To Enhance Reactivity and Selectivity in an Ultrafast Photoclick Reaction.

Light-induced 9,10-phenanthrenequinone-electron-rich alkene (PQ-ERA) photocycloadditions are an attractive new type of photoclick reaction, featuring fast conversions and high biocompatibility. However, the tunability of the reaction was hardly investigated up to now. To this end, we explored the influence of substituents on both reaction partners and the reaction rate between the PQs and ERAs. We identified new handles for functionalization and discovered that using enamines as ERAs leads to drastically enhanced rates (>5400 times faster), high photoreaction quantum yields (ΦP , up to 65 %), and multicolor emission output as well as a high fluorescence quantum yield of the adducts (ΦF , up to 97 %). Further investigation of the photophysical and photochemical properties provided insights to design orthogonal reaction systems both in solution and on nanoparticle surfaces for ultrafast chemoselective functionalization by photoclick reactions.