Prediction Of Structures Research Articles

Design of chiral plasmonic metamaterials based on interpretable deep learning

Chiral plasmonic metamaterials can amplify chiral signals, resulting in circular dichroism (CD) responses that are several orders of magnitude far beyond those of nature. However, the design process of chiral plasmonic metamaterials based on conventional methods is time-consuming. In recent years, the combination of deep learning (DL) and nanophotonics have accelerated the design of nanophotonic devices. Here, we construct the fully connected neural network model for the forward prediction and inverse design of chiral plasmonic metamaterials structures and introduce the permutation importance approach to optimize the model and increase its interpretability. Our experimental results show that using the peak magnitude of CD and the corresponding wavelength instead of the entire spectrum as the output in the forward prediction improves the accuracy of the peak magnitude of CD prediction, avoids the introduction of auxiliary networks, and simplifies the network structure; the permutation importance analysis shows that the gold length of the resonator is the most critical structural parameter affecting the CD response. In the inverse design, the permutation importance method helps us to make feature selections for the input of the network. By reducing 251 inputs (the whole CD spectrum) to 4 inputs (the peak magnitude of CD and the corresponding wavelength), the network can still maintain a good prediction performance and decrease the training time of the network. Our proposed method can be extended not only to other DL models to study the CD response of chiral metamaterials but also to other areas where DL is combined with metamaterials to accelerate the system optimization and design process of nanophotonic devices.

Assessment of three-dimensional RNA structure prediction in CASP15.

The prediction of RNA three-dimensional structures remains an unsolved problem. Here, we report assessments of RNA structure predictions in CASP15, the first CASP exercise that involved RNA structure modeling. Forty-two predictor groups submitted models for at least one of twelve RNA-containing targets. These models were evaluated by the RNA-Puzzles organizers and, separately, by a CASP-recruited team using metrics (GDT, lDDT) and approaches (Z-score rankings) initially developed for assessment of proteins and generalized here for RNA assessment. The two assessments independently ranked the same predictor groups as first (AIchemy_RNA2), second (Chen), and third (RNAPolis and GeneSilico, tied); predictions from deep learning approaches were significantly worse than these top ranked groups, which did not use deep learning. Further analyses based on direct comparison of predicted models to cryogenic electron microscopy (cryo-EM) maps and x-ray diffraction data support these rankings. With the exception of two RNA-protein complexes, models submitted by CASP15 groups correctly predicted the global fold of the RNA targets. Comparisons of CASP15 submissions to designed RNA nanostructures as well as molecular replacement trials highlight the potential utility of current RNA modeling approaches for RNA nanotechnology and structural biology, respectively. Nevertheless, challenges remain in modeling fine details such as noncanonical pairs, in ranking among submitted models, and in prediction of multiple structures resolved by cryo-EM or crystallography.

Development of anti-PD-L1 antibody based on structure prediction of AlphaFold2.

Accurate structural information plays a crucial role in comprehending biological processes and designing drugs. Indeed, the remarkable precision of the AlphaFold2 has facilitated significant advancements in predicting molecular structures, encompassing antibodies and antigens. This breakthrough has paved the way for rational drug design, ushering in new possibilities in the field of pharmaceutical development. Within this study, performing analysis and humanization guided by the structures predicted by AlphaFold2. Notably, the resulting humanized antibody, h3D5-hIgG1, demonstrated exceptional binding affinity to the PD-L1 protein. The KD value of parental antibody 3D5-hIgG1 was increased by nearly 7 times after humanization. Both h3D5-hIgG1 and 3D5-hIgG1 bound to cells expressing human PD-L1 with EC50 values of 5.13 and 9.92nM, respectively. Humanization resulted in a twofold increase in the binding capacity of the antibody, with h3D5-hIgG1 exhibiting superior performance compared to the parental antibody 3D5-hIgG1. Furthermore, h3D5-hIgG1 promoted cytokine secretion of T cells, and significantly suppressed MC38-hPD-L1 tumor growth. This study highlights the potential for artificial intelligence-assisted drug development, which is poised to become a prominent trend in the future.

Progressive collapse resistance prediction of truss string structures under key chord failure

Cables and lower chords at the support play a crucial role in the resistance of truss string structures (TSSs) against progressive collapse. However, the available studies have predominantly focused on cable failure, with seldom research on the progressive collapse mechanism analysis and resistance prediction under the failure of the lower chord at the support. In this study, the accuracy of the finite element (FE) model is verified based on the test results of the TSS exposed to cable failure. Secondly, the criterion for progressive collapse failure of TSSs is given, namely, the structure is considered to have suffered progressive collapse failure when the vertical displacement in the mid-span exceeds 1/60 of the structural span or local collapse occurs. Then, the mechanisms of progressive collapse failure of TSSs exposed to the failure of the lower chord at the support are outlined, and the results show that after the failure of the lower chord at the support, the load paths transmitting the support and the cable are blocked, and the support and the mid-span become the weakest locations of the damaged structure. Moreover, a simplified prediction equation for the progressive collapse resistance of TSSs exposed to the failure of the lower chord at the support is proposed in accordance with some assumptions and simplifications. The comparison reveals that there exists no >20% error between the predicted and FE values, which indicates that the equation has good accuracy and reliability. Therefore, the equation provides a certain reference value for engineering applications.

In vivo-like nearest neighbor parameters improve prediction of fractional RNA base-pairing in cells.

We conducted a thermodynamic analysis of RNA stability in Eco80 artificial cytoplasm, which mimics in vivo conditions, and compared it to transcriptome-wide probing of mRNA. Eco80 contains 80% of Escherichia coli metabolites, with biological concentrations of metal ions, including 2 mM free Mg2+ and 29 mM metabolite-chelated Mg2+. Fluorescence-detected binding isotherms (FDBI) were used to conduct a thermodynamic analysis of 24 RNA helices and found that these helices, which have an average stability of -12.3 kcal/mol, are less stable by ΔΔGo37 ∼1 kcal/mol. The FDBI data was used to determine a set of Watson-Crick free energy nearestneighbor parameters (NNPs), which revealed that Eco80 reduces the stability of three NNPs. This information was used to adjust the NN model using the RNAstructure package. The in vivo-like adjustments have minimal effects on the prediction of RNA secondary structures determined in vitro and in silico, but markedly improve prediction of fractional RNA base pairing in E. coli, as benchmarked with our in vivo DMS and EDC RNA chemical probing data. In summary, our thermodynamic and chemical probing analyses of RNA helices indicate that RNA secondary structures are less stable in cells than in artificially stable in vitro buffer conditions.

Unbranched rod-like RNA is required for RNA editing of hepatitis delta virus genotype 2 and genotype 4

RNA editing of the hepatitis delta virus (HDV) is essential for generating the large delta antigen, which is crucial for virion assembly. In HDV genotype 1 (HDV-1), editing occurs within the context of the unbranched rod-like structure characteristic of HDV RNA, while RNA editing in HDV-3 requires a branched double-hairpin structure. The regulation of RNA editing in HDV-2 and HDV-4 remains uncertain. Based on predictions of the unbranched rod-like RNA structures of HDV-2 and HDV-4, the editing site occurs as an A.C mismatch pair, surrounded by four base pairs upstream and two base pairs downstream of the editing site, respectively. To investigate HDV-2 and HDV-4 RNA editing, cultured cells were transfected with non-replicating editing reporters carrying wild-type sequences or specific mutations. The results revealed that the editing rates observed for wild-type HDV-2 and HDV-4 were fairly similar, albeit lower than that of HDV-1. Like HDV-1, both HDV-2 and HDV-4 showed a reduction in editing rate when the A.C mismatch pair and the immediately upstream base-paired region were disturbed. Notably, extending the downstream base-paired region from two to three or four (forming a structure identical to that of HDV-1) base pairs increased editing rate. Furthermore, we presented novel evidence that indicates the importance of the first bulge's size, located upstream of the editing site, and the base-pairing length within 7–13 and 28–39 nucleotides downstream of the editing site in influencing the HDV-4 editing rate. To summarize, our analyses suggest that the unbranched rod-like structures surrounding the editing site of HDV-2 and HDV-4 play a crucial role in regulating their RNA editing rates.

Accurate Prediction of HSE06 Band Structures for a Diverse Set of Materials Using Δ-Learning

Accurate Prediction of HSE06 Band Structures for a Diverse Set of Materials Using Δ-Learning

Performance prediction of different BCC lattice structures under static loading: an experimental approach

Performance prediction of different BCC lattice structures under static loading: an experimental approach

Naive Prediction of Protein Backbone Phi and Psi Dihedral Angles Using Deep Learning.

Protein structure prediction represents a significant challenge in the field of bioinformatics, with the prediction of protein structures using backbone dihedral angles recently achieving significant progress due to the rise of deep neural network research. However, there is a trend in protein structure prediction research to employ increasingly complex neural networks and contributions from multiple models. This study, on the other hand, explores how a single model transparently behaves using sequence data only and what can be expected from the predicted angles. To this end, the current paper presents data acquisition, deep learning model definition, and training toward the final protein backbone angle prediction. The method applies a simple fully connected neural network (FCNN) model that takes only the primary structure of the protein with a sliding window of size 21 as input to predict protein backbone ϕ and ψ dihedral angles. Despite its simplicity, the model shows surprising accuracy for the ϕ angle prediction and somewhat lower accuracy for the ψ angle prediction. Moreover, this study demonstrates that protein secondary structure prediction is also possible with simple neural networks that take in only the protein amino-acid residue sequence, but more complex models are required for higher accuracies.

Transferable equivariant graph neural networks for the Hamiltonians of molecules and solids

This work presents an E(3) equivariant graph neural network called HamGNN, which can fit the electronic Hamiltonian matrix of molecules and solids by a complete data-driven method. Unlike invariant models that achieve equivariance approximately through data augmentation, HamGNN employs E(3) equivariant convolutions to construct the Hamiltonian matrix, ensuring strict adherence to all equivariant constraints inherent in the physical system. In contrast to previous models with limited transferability, HamGNN demonstrates exceptional accuracy on various datasets, including QM9 molecular datasets, carbon allotropes, silicon allotropes, SiO2 isomers, and BixSey compounds. The trained HamGNN models exhibit accurate predictions of electronic structures for large crystals beyond the training set, including the Moiré twisted bilayer MoS2 and silicon supercells with dislocation defects, showcasing remarkable transferability and generalization capabilities. The HamGNN model, trained on small systems, can serve as an efficient alternative to density functional theory (DFT) for accurately computing the electronic structures of large systems.

Toward automatic generation of control structures for process flow diagrams with large language models

AbstractDeveloping Piping and Instrumentation Diagrams (P&IDs) is a crucial step during process development. We propose a data‐driven method for the prediction of control structures. Our methodology is inspired by end‐to‐end transformer‐based human language translation models. We cast the control structure prediction as a translation task where Process Flow Diagrams (PFDs) without control structures are translated to PFDs with control structures. We represent the topology of PFDs as strings using the SFILES 2.0 notation. We pretrain our model using generated PFDs to learn the grammatical structure. Thereafter, the model is fine‐tuned leveraging transfer learning on real PFDs. The model achieved a top‐5 accuracy of 74.8% on 10,000 generated PFDs and 89.2% on 100,000 generated PFDs. These promising results show great potential for AI‐assisted process engineering. The tests on a dataset of 312 real PFDs indicate the need for a larger PFD dataset for industry applications and hybrid artificial intelligence solutions.

Structural and molecular biology of Sabiá virus.

Brazilian mammarenavirus, or Sabiá virus (SABV), is a New World (NW) arenavirus associated with fulminant hemorrhagic disease in humans and the sole biosafety level 4 microorganism ever isolated in Brazil. Since the isolation of SABV in the 1990s, studies on viral biology have been scarce, with no available countermeasures against SABV infection or disease. Here we provide a comprehensive review of SABV biology, including key aspects of SABV replication, and comparisons with related Old World and NW arenaviruses. SABV is most likely a rodent-borne virus, transmitted to humans, through exposure to urine and feces in peri-urban areas. Using protein structure prediction methods and alignments, we analyzed shared and unique features of SABV proteins (GPC, NP, Z, and L) that could be explored in search of therapeutic strategies, including repurposing intended application against arenaviruses. Highly conserved catalytic activities present in L protein could be targeted for broad-acting antiviral activity among arenaviruses, while protein-protein interactions, such as those between L and the matrix protein Z, have evolved in NW arenaviruses and should be specific to SABV. The nucleoprotein (NP) also shares targetable interaction interfaces with L and Z and exhibits exonuclease activity in the C-terminal domain, which may be involved in multiple aspects of SABV replication. Envelope glycoproteins GP1 and GP2 have been explored in the development of promising cross-reactive neutralizing antibodies and vaccines, some of which could be repurposed for SABV. GP1 remains a challenging target in SABV as evolutive pressures render it the most variable viral protein in terms of both sequence and structure, while antiviral strategies targeting the Z protein remain to be validated. In conclusion, the prediction and analysis of protein structures should revolutionize research on viruses such as SABV by facilitating the rational design of countermeasures while reducing dependence on sophisticated laboratory infrastructure for experimental validation.

GA-Net: A geographical attention neural network for the segmentation of body torso tissue composition

PurposeBody composition analysis (BCA) of the body torso plays a vital role in the study of physical health and pathology and provides biomarkers that facilitate the diagnosis and treatment of many diseases, such as type 2 diabetes mellitus, cardiovascular disease, obstructive sleep apnea, and osteoarthritis. In this work, we propose a body composition tissue segmentation method that can automatically delineate those key tissues, including subcutaneous adipose tissue, skeleton, skeletal muscle tissue, and visceral adipose tissue, on positron emission tomography/computed tomography scans of the body torso. MethodsTo provide appropriate and precise semantic and spatial information that is strongly related to body composition tissues for the deep neural network, first we introduce a new concept of the body area and integrate it into our proposed segmentation network called Geographical Attention Network (GA-Net). The body areas are defined following anatomical principles such that the whole body torso region is partitioned into three non-overlapping body areas. Each body composition tissue of interest is fully contained in exactly one specific minimal body area. Secondly, the proposed GA-Net has a novel dual-decoder schema that is composed of a tissue decoder and an area decoder. The tissue decoder segments the body composition tissues, while the area decoder segments the body areas as an auxiliary task. The features of body areas and body composition tissues are fused through a soft attention mechanism to gain geographical attention relevant to the body tissues. Thirdly, we propose a body composition tissue annotation approach that takes the body area labels as the region of interest, which significantly improves the reproducibility, precision, and efficiency of delineating body composition tissues. ResultsOur evaluations on 50 low-dose unenhanced CT images indicate that GA-Net outperforms other architectures statistically significantly based on the Dice metric. GA-Net also shows improvements for the 95% Hausdorff Distance metric in most comparisons. Notably, GA-Net exhibits more sensitivity to subtle boundary information and produces more reliable and robust predictions for such structures, which are the most challenging parts to manually mend in practice, with potentially significant time-savings in the post hoc correction of these subtle boundary placement errors. Due to the prior knowledge provided from body areas, GA-Net achieves competitive performance with less training data. Our extension of the dual-decoder schema to TransUNet and 3D U-Net demonstrates that the new schema significantly improves the performance of these classical neural networks as well. Heatmaps obtained from attention gate layers further illustrate the geographical guidance function of body areas for identifying body tissues. Conclusions(i) Prior anatomic knowledge supplied in the form of appropriately designed anatomic container objects significantly improves the segmentation of bodily tissues. (ii) Of particular note are the improvements achieved in the delineation of subtle boundary features which otherwise would take much effort for manual correction. (iii) The method can be easily extended to existing networks to improve their accuracy for this application.

Expeditious Discovery of Small-Molecule Thermoresponsive Ionic Liquid Materials: A Review.

Ionic liquids (ILs) are a class of low-melting molten salts (<100 °C) constituted entirely of ions, and their research has gained tremendous attention in line with their remarkably growing applications (>124,000 publications dated 30 August 2023 from the Web of ScienceTM). In this review, we first briefly discussed the recent developments and unique characteristics of ILs and zwitterionic liquids (ZILs). Compared to molecular solvents and other conventional organic compounds, (zwitter) ionic liquids carry negligible volatility and are potentially recyclable and reusable. For structures, both ILs and ZILs can be systematically tailor-designed and engineered and are synthetically fine-tunable. As such, ionic liquids, including chiral, supported, task-specific ILs, have been widely used as powerful ionic solvents as well as valuable additives and catalysts for many chemical reactions. Moreover, ILs have demonstrated their value for use as polymerase chain reaction (PCR) enhancers for DNA amplification, chemoselective artificial olfaction for targeted VOC analysis, and recognition-based affinity extraction. As the major focus of this review, we extensively discussed that small-molecule thermoresponsive ILs (TILs) and ZILs (zwitterionic TILs) are new types of smart materials and can be expeditiously discovered through the structure and phase separation (SPS) relationship study by the combinatorial approach. Using this SPS platform developed in our laboratory, we first depicted the rapid discovery of N,N-dialkylcycloammonium and 1,3,4-trialkyl-1,2,3-triazolium TILs that concomitantly exhibited LCST (lower critical solution temperature) phase transition in water and displayed biochemically attractive Tc values. Both smart IL materials were suited for applications to proteins and other biomolecules. Zwitterionic TILs are ZILs whose cations and anions are tethered together covalently and are thermoresponsive to temperature changes. These zwitterionic TIL materials can serve as excellent extraction solvents, through temperature change, for biomolecules such as proteins since they differ from the common TIL problems often associated with unwanted ion exchanges during extractions. These unique structural characteristics of zwitterionic TIL materials greatly reduce and may avoid the denaturation of proteins under physiological conditions. Lastly, we argued that both rational structural design and combinatorial library synthesis of small-molecule TIL materials should take into consideration the important issues of their cytotoxicity and biosafety to the ecosystem, potentially causing harm to the environment and directly endangering human health. Finally, we would concur that before precise prediction and quantitative simulation of TIL structures can be realized, combinatorial chemistry may be the most convenient and effective technology platform to discover TIL expeditiously. Through our rational TIL design and combinatorial library synthesis and screening, we have demonstrated its power to discover novel chemical structures of both TILs and zwitterionic TILs. Undoubtedly, we will continue developing new small-molecule TIL structures and studying their applications related to other thermoresponsive materials.

Low-Temperature Predicted Structures of Ag2S (Silver Sulfide).

Silver sulfide phases, such as body-centered cubic argentite and monoclinic acanthite, are widely known. Traditionally, acanthite is regarded as the only low-temperature phase of silver sulfide. However, the possible existence of other low-temperature phases of silver sulfide cannot be ruled out. Until now, there have been only a few suggestions about low-temperature Ag2S phases that differ from monoclinic acanthite. The lack of a uniform approach has hampered the prediction of such phases. In this work, the use of such an effective tool as an evolutionary algorithm for the first time made it possible to perform a broad search for the model Ag2S phases of silver sulfide, which are low-temperature with respect to cubic argentite. The possibility of forming Ag2S phases with cubic, tetragonal, orthorhombic, trigonal, monoclinic, and triclinic symmetry is considered. The calculation of the cohesion energy and the formation enthalpy show, for the first time, that the formation of low-symmetry Ag2S phases is energetically most favorable. The elastic stiffness constants cij of all predicted Ag2S phases are computed, and their mechanical stability is determined. The densities of the electronic states of the predicted Ag2S phases are calculated. The prediction of low-temperature Ag2S structures indicates the possibility of synthesizing new silver sulfide phases with improved properties.

Extracting Structural Information from Physicochemical Property Measurements Using Machine Learning─A New Approach for Structure Elucidation in Non-targeted Analysis.

Non-targeted analysis (NTA) has made critical contributions in the fields of environmental chemistry and environmental health. One critical bottleneck is the lack of available analytical standards for most chemicals in the environment. Our study aims to explore a novel approach that integrates measurements of equilibrium partition ratios between organic solvents and water (KSW) to predictions of molecular structures. These properties can be used as a fingerprint, which with the help of a machine learning algorithm can be converted into a series of functional groups (RDKit fragments), which can be used to search chemical databases. We conducted partitioning experiments using a chemical mixture containing 185 chemicals in 10 different organic solvents and water. Both a liquid chromatography quadrupole time-of-flight mass spectrometer (LC-QTOF MS) and a LC-Orbitrap MS were used to assess the feasibility of the experimental method and the accuracy of the algorithm at predicting the correct functional groups. The two methods showed differences in log KSW with the QTOF method showing a mean absolute error (MAE) of 0.22 and the Orbitrap method 0.33. The differences also culminated into errors in the predictions of RDKit fragments with the MAE for the QTOF method being 0.23 and for the Orbitrap method being 0.31. Our approach presents a new angle in structure elucidation for NTA and showed promise in assisting with compound identification.

MLPER: A Machine Learning-Based Prediction Model for Building Earthquake Response Using Ambient Vibration Measurements

Deep neural networks (DNNs) have gained prominence in addressing regression problems, offering versatile architectural designs that cater to various applications. In the field of earthquake engineering, seismic response prediction is a critical area of study. Simplified models such as single-degree-of-freedom (SDOF) and multi-degree-of-freedom (MDOF) systems have traditionally provided valuable insights into structural behavior, known for their computational efficiency facilitating faster simulations. However, these models have notable limitations in capturing the nuanced nonlinear behavior of structures and the spatial variability of ground motions. This study focuses on leveraging ambient vibration (AV) measurements of buildings, combined with earthquake (EQ) time-history data, to create a predictive model using a neural network (NN) in image format. The primary objective is to predict a specific building’s earthquake response accurately. The training dataset consists of 1197 MDOF 2D shear models, generating a total of 32,319 training samples. To evaluate the performance of the proposed model, termed MLPER (machine learning-based prediction of building structures’ earthquake response), several metrics are employed. These include the mean absolute percentage error (MAPE) and the mean deviation angle (MDA) for comparisons in the time domain. Additionally, we assess magnitude-squared coherence values and phase differences (Δφ) for comparisons in the frequency domain. This study underscores the potential of the MLPER as a reliable tool for predicting building earthquake responses, addressing the limitations of simplified models. By integrating AV measurements and EQ time-history data into a neural network framework, the MLPER offers a promising avenue for enhancing our understanding of structural behavior during seismic events, ultimately contributing to improved earthquake resilience in building design and engineering.

Low cycle fatigue lifetime prediction of superplastic shape memory alloy structures: Application to endodontic instruments

In this paper we propose a methodology for a fast numerical determination of low cycle fatigue lifetime of superelastic shape memory alloy structures. This method is based on the observation that generally, in low cycle fatigue, shape memory alloy (SMA) structures are subject to loadings that lead to a confined non-linear behaviour at stress concentration points, such as notches. Numerical fatigue lifetime prediction requires the computation of the mechanical state at critical points. However, classical computational methods, like the non-linear finite element method, lead to a prohibitive computation time in a non-linear cyclic framework. To overcome this issue, we propose to use fast prediction methods, based on localization laws. Following the determination of the stabilized behaviour, an energetic fatigue criterion is applied. The numerical fatigue life prediction model is validated experimentally on SMA endodontic instruments.

Accurately identifying nucleic-acid-binding sites through geometric graph learning on language model predicted structures.

The interactions between nucleic acids and proteins are important in diverse biological processes. The high-quality prediction of nucleic-acid-binding sites continues to pose a significant challenge. Presently, the predictive efficacy of sequence-based methods is constrained by their exclusive consideration of sequence context information, whereas structure-based methods are unsuitable for proteins lacking known tertiary structures. Though protein structures predicted by AlphaFold2 could be used, the extensive computing requirement of AlphaFold2 hinders its use for genome-wide applications. Based on the recent breakthrough of ESMFold for fast prediction of protein structures, we have developed GLMSite, which accurately identifies DNA- and RNA-binding sites using geometric graph learning on ESMFold predicted structures. Here, the predicted protein structures are employed to construct protein structural graph with residues as nodes and spatially neighboring residue pairs for edges. The node representations are further enhanced through the pre-trained language model ProtTrans. The network was trained using a geometric vector perceptron, and the geometric embeddings were subsequently fed into a common network to acquire common binding characteristics. Finally, these characteristics were input into two fully connected layers to predict binding sites with DNA and RNA, respectively. Through comprehensive tests on DNA/RNA benchmark datasets, GLMSite was shown to surpass the latest sequence-based methods and be comparable with structure-based methods. Moreover, the prediction was shown useful for inferring nucleic-acid-binding proteins, demonstrating its potential for protein function discovery. The datasets, codes, and trained models are available at https://github.com/biomed-AI/nucleic-acid-binding.

Diverse ranking metamaterial inverse design based on contrastive and transfer learning.

Metamaterials, thoughtfully designed, have demonstrated remarkable success in the manipulation of electromagnetic waves. More recently, deep learning can advance the performance in the field of metamaterial inverse design. However, existing inverse design methods based on deep learning often overlook potential trade-offs of optimal design and outcome diversity. To address this issue, in this work we introduce contrastive learning to implement a simple but effective global ranking inverse design framework. Viewing inverse design as spectrum-guided ranking of the candidate structures, our method creates a resemblance relationship of the optical response and metamaterials, enabling the prediction of diverse structures of metamaterials based on the global ranking. Furthermore, we have combined transfer learning to enrich our framework, not limited in prediction of single metamaterial representation. Our work can offer inverse design evaluation and diverse outcomes. The proposed method may shrink the gap between flexibility and accuracy of on-demand design.