Molecular Feature Extraction Research Articles

MREDTA: A BERT and transformer-based molecular representation encoder for predicting drug-target binding affinity.

Drug-target binding affinity (DTA) prediction is vital for drug repositioning. The accuracy and generalizability of DTA models remain a major challenge. Here, we develop a model composed of BERT-Trans Block, Multi-Trans Block, and DTI Learning modules, referred to as Molecular Representation Encoder-based DTA prediction (MREDTA). MREDTA has three advantages: (1) extraction of both local and global molecular features simultaneously through skip connections; (2) improved sensitivity to molecular structures through the Multi-Trans Block; (3) enhanced generalizability through the introduction of BERT. Compared with 12 advanced models, benchmark testing of KIBA and Davis datasets demonstrated optimal performance of MREDTA. In case study, we applied MREDTA to 2034 FDA-approved drugs for treating non-small-cell lung cancer (NSCLC), all of which act on mutant EGFRT790M protein. The corresponding molecular docking results demonstrated the robustness of MREDTA.

MDs-NP: a property prediction model construction procedure for naphtha based on molecular dynamics simulation

In the context of carbon neutrality and carbon peaking, molecular management has become a focus of the petrochemical industry. The key to achieving molecular management is molecular reconstruction, which relies on rapid and accurate calculation of oil properties. Focusing on naphtha, we proposed a novel property prediction model construction procedure (MDs-NP) employing molecular dynamics simulations for property collections and gamma distribution from real analytical data for calculating mole fractions of simulation mixtures. We calculated 348 sets of mixture properties data in the range of 273 K–300 K by molecular dynamics simulations. Molecular feature extraction was based on molecular descriptors. In addition to descriptors based on open-source toolkits (RDKit and Mordred), we designed 12 naphtha knowledge (NK) descriptors with a focus on naphtha. Three machine learning algorithms (support vector regression, extreme gradient boosting and artificial neural network) were applied and compared to establish models for the prediction of the density and viscosity of naphtha. Mordred and NK descriptors + support vector regression algorithm achieved the best performance for density. The selected RDKFp and NK descriptors + artificial neural network algorithm achieved the best performance for viscosity. Using ablation studies, T, P_w and CC(C)C are three effective descriptors in NK that can improve the performance of the property prediction models. MDs-NP has the potential to be extended to more properties as well as more-complex petroleum systems. The models from MDs-NP can be used for rapid molecular reconstruction to facilitate construction of data-driven models and intelligent transformation of petrochemical processes.

Triple Generative Self-Supervised Learning Method for Molecular Property Prediction.

Molecular property prediction is an important task in drug discovery, and with help of self-supervised learning methods, the performance of molecular property prediction could be improved by utilizing large-scale unlabeled dataset. In this paper, we propose a triple generative self-supervised learning method for molecular property prediction, called TGSS. Three encoders including a bi-directional long short-term memory recurrent neural network (BiLSTM), a Transformer, and a graph attention network (GAT) are used in pre-training the model using molecular sequence and graph structure data to extract molecular features. The variational auto encoder (VAE) is used for reconstructing features from the three models. In the downstream task, in order to balance the information between different molecular features, a feature fusion module is added to assign different weights to each feature. In addition, to improve the interpretability of the model, atomic similarity heat maps were introduced to demonstrate the effectiveness and rationality of molecular feature extraction. We demonstrate the accuracy of the proposed method on chemical and biological benchmark datasets by comparative experiments.

Application of clustering strategy for automatic segmentation of tissue regions in mass spectrometry imaging.

Mass spectrometry imaging (MSI) has been widely used in biomedical research fields. Each pixel in MSI consists of a mass spectrum that reflects the molecule feature of the tissue spot. Because MSI contains high-dimensional datasets, it is highly desired to develop computational methods for data mining and constructing tissue segmentation maps. To visualize different tissue regions based on mass spectrum features and improve the efficiency in processing enormous data, we proposed a computational strategy that consists of four procedures including preprocessing, data reduction, clustering, and quantitative validation. In this study, we examined the combination of t-distributed stochastic neighbor embedding (t-SNE) and hierarchical clustering (HC) for MSI data analysis. Using publicly available MSI datasets, one dataset of mouse urinary bladder, and one dataset of human colorectal cancer, we demonstrated that the generated tissue segmentation maps from this combination were superior to other data reduction and clustering algorithms. Using the staining image as a reference, we assessed the performance of clustering algorithms with external and internal clustering validation measures, including purity, adjusted Rand index (ARI), Davies-Bouldin index (DBI), and spatial aggregation index (SAI). The result indicated that SAI delivered excellent performance for automatic segmentation of tissue regions in MSI. We used a clustering algorithm to construct tissue automatic segmentation in MSI datasets. The performance was evaluated by comparing it with the stained image and calculating clustering validation indexes. The results indicated that SAI is important for automatic tissue segmentation in MSI, different from traditional clustering validation measures. Compared to the reports that used internal clustering validation measures such as DBI, our method offers more effective evaluation of clustering results for MSI segmentation. We envision that the proposed automatic image segmentation strategy can facilitate deep learning in molecular feature extraction and biomarker discovery for the biomedical applications of MSI.

Deep learning-based classification model for GPR151 activator activity prediction

BackgroundGPR151 is a kind of protein belonging to G protein-coupled receptor family that is closely associated with a variety of physiological and pathological processes.The potential use of GPR151 as a therapeutic target for the management of metabolic disorders has been demonstrated in several studies, highlighting the demand to explore its activators further. Activity prediction serves as a vital preliminary step in drug discovery, which is both costly and time-consuming. Thus, the development of reliable activity classification model has become an essential way in the process of drug discovery, aiming to enhance the efficiency of virtual screening.ResultsWe propose a learning-based method based on feature extractor and deep neural network to predict the activity of GPR151 activators. We first introduce a new molecular feature extraction algorithm which utilizes the idea of bag-of-words model in natural language to densify the sparse fingerprint vector. Mol2vec method is also used to extract diverse features. Then, we construct three classical feature selection algorithms and three types of deep learning model to enhance the representational capacity of molecules and predict activity label by five different classifiers. We conduct experiments using our own dataset of GPR151 activators. The results demonstrate high classification accuracy and stability, with the optimal model Mol2vec-CNN significantly improving performance across multiple classifiers. The svm classifier achieves the best accuracy of 0.92 and F1 score of 0.76 which indicates promising applications for our method in the field of activity prediction.ConclusionThe results suggest that the experimental design of this study is appropriate and well-conceived. The deep learning-based feature extraction algorithm established in this study outperforms traditional feature selection algorithm for activity prediction. The model developed can be effectively utilized in the pre-screening stage of drug virtual screening.

In silico prediction of hERG blockers using machine learning and deep learning approaches.

The human ether-à-go-go-related gene (hERG) is associated with drug cardiotoxicity. If the hERG channel is blocked, it will lead to prolonged QT interval and cause sudden death in severe cases. Therefore, it is important to evaluate the hERG-blocking property of compounds in early drug discovery. In this study, a dataset containing 4556 compounds with IC50 values determined by patch clamp techniques on mammalian lineage cells was collected, and hERG blockers and non-blockers were distinguished according to 3 single-thresholds and 2 binary thresholds. Four machine learning (ML) algorithms combining four molecular fingerprints and molecular descriptors as well as graph convolutional neural networks (GCN) were used to construct a series of binary classification models. The results showed that the best models varied for different thresholds. The ML models implemented by support vector machine and random forest performed well based on Morgan fingerprints and molecular descriptors, with AUCs ranging from 0.884 to 0.950. GCN showed superior prediction performance with AUCs above 0.952, which might be related to its direct extraction of molecular features from the original input. Meanwhile, the classification of binary threshold was better than that of single threshold, which could provide us with a more accurate prediction of hERG blockers. At last, the applicability domain for the model was defined, and seven structural alerts that might generate hERG blockage were identified by information gain and substructure frequency analysis. Our work would be beneficial for identifying hERG blockers in chemicals.

HIV-1/HBV Coinfection Accurate Multitarget Prediction Using a Graph Neural Network-Based Ensemble Predicting Model.

HIV and HBV infection are both serious public health challenges. There are more than approximately 4 million patients coinfected with HIV and HBV worldwide, and approximately 5% to 15% of those infected with HIV are coinfected with HBV. Disease progression is more rapid in patients with coinfection, which significantly increases the likelihood of patients progressing from chronic hepatitis to cirrhosis, end-stage liver disease, and hepatocellular carcinoma. HIV treatment is complicated by drug interactions, antiretroviral (ARV) hepatotoxicity, and HBV-related immune reconditioning and inflammatory syndromes. Drug development is a highly costly and time-consuming procedure with traditional experimental methods. With the development of computer-aided drug design techniques, both machine learning and deep learning have been successfully used to facilitate rapid innovations in the virtual screening of candidate drugs. In this study, we proposed a graph neural network-based molecular feature extraction model by integrating one optimal supervised learner to replace the output layer of the GNN to accurately predict the potential multitargets of HIV-1/HBV coinfections. The experimental results strongly suggested that DMPNN + GBDT may greatly improve the accuracy of binary-target predictions and efficiently identify the potential multiple targets of HIV-1 and HBV simultaneously.

Real-time measurement of phase partitioning of organic compounds using a proton-transfer-reaction time-of-flight mass spectrometer coupled to a CHARON inlet

Abstract. Understanding the gas–particle partitioning of semivolatile organic compounds (SVOCs) is of crucial importance in the accurate representation of the global budget of atmospheric organic aerosols. In this study, we quantified the gas- vs. particle-phase fractions of a large number of SVOCs in real time in an urban area of East China with the use of a CHemical Analysis of aeRosols ONline (CHARON) inlet coupled to a high-resolution proton-transfer-reaction time-of-flight mass spectrometer (PTR-ToF-MS). We demonstrated the use of the CHARON inlet for highly efficient collection of particulate SVOCs while maintaining the intact molecular structures of these compounds. The collected month-long dataset with hourly resolution allows us to examine the gas–particle partitioning of a variety of SVOCs under ambient conditions. By comparing the measurements with model predictions using instantaneous equilibrium partitioning theory, we found that the dissociation of large parent molecules during the PTR ionization process likely introduces large uncertainties to the measured gas- vs. particle-phase fractions of less oxidized SVOCs, and therefore, caution should be taken when linking the molecular composition to the particle volatility when interpreting the PTR-ToF-MS data. Our analysis suggests that understanding the fragmentation mechanism of SVOCs and accounting for the neutral losses of small moieties during the molecular feature extraction from the raw PTR mass spectra could reduce, to a large extent, the uncertainties associated with the gas–particle partitioning measurement of SVOCs in the ambient atmosphere.

MIFNN: Molecular Information Feature Extraction and Fusion Deep Neural Network for Screening Potential Drugs.

Molecular property prediction is essential for drug screening and reducing the cost of drug discovery. Current approaches combined with deep learning for drug prediction have proven their viability. Based on the previous deep learning networks, we propose the Molecular Information Fusion Neural Network (MIFNN). The features of MIFNN are as follows: (1) we extracted directed molecular information using 1D-CNN and the Morgan fingerprint using 2D-CNN to obtain more comprehensive feature information; (2) we fused two molecular features from one-dimensional and two-dimensional space, and we used the directed message-passing method to reduce the repeated collection of information and improve efficiency; (3) we used a bidirectional long short-term memory and attention module to adjust the molecular feature information and improve classification accuracy; (4) we used the particle swarm optimization algorithm to improve the traditional support vector machine. We tested the performance of the model on eight publicly available datasets. In addition to comparing the overall classification capability with the baseline model, we conducted a series of ablation experiments to verify the optimization of different modules in the model. Compared with the baseline model, our model achieved a maximum improvement of 14% on the ToxCast dataset. The performance was very stable on most datasets. On the basis of the current experimental results, MIFNN performed better than previous models on the datasets applied in this paper.

Accurate Prediction of Epigenetic Multi-Targets with Graph Neural Network-Based Feature Extraction.

Epigenetics, referring to genetic modifications that change gene expression, but which are not encoded in DNA, has been shown to be related to oncology, with the potential to influence associated treatments. As such, epigenetic drugs comprise an important new field in cancer therapy; however, drug development is a high-cost and time-consuming procedure. Different epigenetic modifications, such as mutations in DNA methyltransferase and somatic mutations in core histone genes that lead to a global loss of the histone modifications, have innumerable relationships. In this article, we propose a graph neural network-based model for the extraction of molecular features, thus reducing the computational requirements. Through integration with a popular and efficient supervised learner, our model achieves higher prediction accuracy in both single- and multi-target tasks and can determine the pleiotropy associated with drugs, providing theoretical support for drug combination and discovery research.

Data-driven identification of plasma metabolite clusters and metabolites of interest for potential detection of early-stage non-small cell lung cancer cases versus cancer-free controls

BackgroundMetabolomics is a potential means for biofluid-based lung cancer detection. We conducted a non-targeted, data-driven assessment of plasma from early-stage non-small cell lung cancer (ES-NSCLC) cases versus cancer-free controls (CFC) to explore and identify the classes of metabolites for further targeted metabolomics biomarker development.MethodsPlasma from 250 ES-NSCLC cases and 250 CFCs underwent ultra-high-performance liquid chromatography/quadrupole time-of-flight mass spectrometry (UHPLC-QTOF-MS) in positive and negative electrospray ionization (ESI) modes. Molecular feature extraction, formula generation, and find-by-ion tools annotated metabolic entities. Analysis was restricted to endogenous metabolites present in ≥ 80% of samples. Unsupervised hierarchical cluster analysis identified clusters of metabolites. The metabolites with the strongest correlation with the principal component of each cluster were included in logistic regression modeling to assess discriminatory performance with and without adjustment for clinical covariates.ResultsA total of 1900 UHPLC-QTOF-MS assessments identified 1667 and 2032 endogenous metabolites in the ESI-positive and ESI-negative modes, respectively. After data filtration, 676 metabolites remained, and 12 clusters of metabolites were identified from each ESI mode. Multivariable logistic regression using the representative metabolite from each cluster revealed effective classification of cases from controls with overall diagnostic accuracy of 91% (ESI positive) and 94% (ESI negative). Metabolites of interest identified for further targeted analysis include the following: 1b, 3a, 12a-trihydroxy-5b-cholanoic acid, pyridoxamine 5′-phosphate, sphinganine 1-phosphate, gamma-CEHC, 20-carboxy-leukotriene B4, isodesmosine, and 18-hydroxycortisol.ConclusionsPlasma-based metabolomic detection of early-stage NSCLC appears feasible. Further metabolomics studies targeting phospholipid, steroid, and fatty acid metabolism are warranted to further develop noninvasive metabolomics-based detection of early-stage NSCLC.

Message-passing neural network based multi-task deep-learning framework for COSMO-SAC based σ-profile and VCOSMO prediction

• A message-passing neural network based feedforward neural network (MPNN-FNN) is developed for σ-profile and V COSMO prediction. • Hybrid molecular representations are utilized to extract the molecular features. • The proposed model can well discriminate the cis/trans and structural isomers. • The IDAC calculated by predicted σ -profile and V COSMO is very close to the QM derived ones. The correct implementation of the quantum mechanics (QM) calculation for COSMO-SAC based surface charge density profiles ( σ -profile) and cavity volumes ( V COSMO ) is tricky and time-consuming. As a shortcut, the GC approaches (GC-COSMO) were proposed to predict the required σ -profile and V COSMO for COSMO based methods. However, the GC-COSMO model cannot discriminate the isomers and cannot predict the complex molecules due to a lack of diverse functional groups. In this context, a message-passing neural network based feedforward neural network (MPNN-FNN) is proposed for rapid and accurate predicting COSMO-SAC based σ -profile and V COSMO . The hybrid molecular representation integrated the message-passing neural network (MPNN) learned representations and 200 additional molecule-level RDKit features, which can capture both local and global features of the overall molecule, is utilized to develop a better molecular feature extraction model. Based on the hybrid molecular representations, a feedforward neural network (FNN) model for the σ -profile and V COSMO prediction is trained and tested on the VT-2005 database. The developed MPNN-FNN model demonstrates significantly better performance than the GC-COSMO method and can well discriminate the cis/trans and structural isomers. In addition, the calculation speed based on the proposed model for COSMO-SAC based infinite dilution activity coefficient (IDAC) reaches about 100 compounds per second, which are the prerequisite to performing a high throughput green solvent screening.

Molecular pathological recognition of freshly excised human glioma using terahertz ATR spectroscopy.

The diagnosis and treatment of glioma depends greatly on the rapid extraction of molecular pathological features. In this study, human brain tumor tissues of different grades were analyzed using terahertz (THz) attenuated total reflectance (ATR) time-domain spectroscopy. Substantial differences in THz parameters were observed between paracarcinoma tissue and grade I-IV gliomas, Furthermore, the difference of THz absorption coefficient increases with the increase of THz frequency. It was also demonstrated that the isocitrate dehydrogenase (IDH) mutant and wild-type glioma tissues can be well distinguished using THz spectroscopy. Therefore, THz ATR spectroscopy can realize molecular typing recognition based on molecular pathology. This will provide a theoretical basis for developing intraoperative real-time glioma recognition and diagnosis technology.

MStractor: R Workflow Package for Enhancing Metabolomics Data Pre-Processing and Visualization.

Untargeted metabolomics experiments for characterizing complex biological samples, conducted with chromatography/mass spectrometry technology, generate large datasets containing very complex and highly variable information. Many data-processing options are available, however, both commercial and open-source solutions for data processing have limitations, such as vendor platform exclusivity and/or requiring familiarity with diverse programming languages. Data processing of untargeted metabolite data is a particular problem for laboratories that specialize in non-routine mass spectrometry analysis of diverse sample types across humans, animals, plants, fungi, and microorganisms. Here, we present MStractor, an R workflow package developed to streamline and enhance pre-processing of metabolomics mass spectrometry data and visualization. MStractor combines functions for molecular feature extraction with user-friendly dedicated GUIs for chromatographic and mass spectromerty (MS) parameter input, graphical quality-control outputs, and descriptive statistics. MStractor performance was evaluated through a detailed comparison with XCMS Online. The MStractor package is freely available on GitHub at the MetabolomicsSA repository.

XGraphBoost: Extracting Graph Neural Network-Based Features for a Better Prediction of Molecular Properties.

Determining the properties of chemical molecules is essential for screening candidates similar to a specific drug. These candidate molecules are further evaluated for their target binding affinities, side effects, target missing probabilities, etc. Conventional machine learning algorithms demonstrated satisfying prediction accuracies of molecular properties. A molecule cannot be directly loaded into a machine learning model, and a set of engineered features needs to be designed and calculated from a molecule. Such hand-crafted features rely heavily on the experiences of the investigating researchers. The concept of graph neural networks (GNNs) was recently introduced to describe the chemical molecules. The features may be automatically and objectively extracted from the molecules through various types of GNNs, e.g., GCN (graph convolution network), GGNN (gated graph neural network), DMPNN (directed message passing neural network), etc. However, the training of a stable GNN model requires a huge number of training samples and a large amount of computing power, compared with the conventional machine learning strategies. This study proposed the integrated framework XGraphBoost to extract the features using a GNN and build an accurate prediction model of molecular properties using the classifier XGBoost. The proposed framework XGraphBoost fully inherits the merits of the GNN-based automatic molecular feature extraction and XGBoost-based accurate prediction performance. Both classification and regression problems were evaluated using the framework XGraphBoost. The experimental results strongly suggest that XGraphBoost may facilitate the efficient and accurate predictions of various molecular properties. The source code is freely available to academic users at https://github.com/chenxiaowei-vincent/XGraphBoost.git.

Comprehensive Characterization of Bovine Milk Lipids: Triglycerides.

A comprehensive survey on triglycerides (TAGs) of bovine milk was conducted by a combination of exhaustive liquid chromatography (LC) separation, high-resolution mass spectrometry (MS) detection, and automated lipid molecular feature extraction. A total of 220 groups (a series of species having the same chemical formula and mass) and 3454 molecular species of TAGs were identified based on the accurate mass of the parent ion as well as MS2 information. Sixty-five different fatty acids (FAs) were found across these TAG species; C4:0, C6:0, C8:0, C10:0, C12:0, C14:0, C16:0, C18:0, and C18:1 were the most frequent FAs, whereas C11:3, C11:4, C27:0, C27:1, C28:0, and C28:1 were rare FAs in TAG molecules. The number of species identified represents only a small portion of total TAG molecules that can be theoretically synthesized from 65 FAs. Each TAG group contains on average 15–16 isomeric species (species with different FA composition), but positional isomers do not seem to be widespread. As the isomeric species cannot be completely resolved chromatographically, quantification of TAG was conducted at the group level. The most abundant TAG groups in bovine milk include TAG 34:0, TAG 36:0, TAG 38:1, TAG 38:0, and TAG 40:1. This study provides the most comprehensive milk TAG inventory so far that can be used as a reference for studying milk lipids.

A novel unambiguous strategy of molecular feature extraction in machine learning assisted predictive models for environmental properties

A novel strategy of molecular feature extraction coupling with the machine learning algorithm for developing predictive models of environmental properties.

Extraction of molecular features for the drug discovery targeting protein-protein interaction of Helicobacter pylori CagA and tumor suppressor protein ASSP2.

Half of the world population is infected by the Gram-negative bacterium Helicobacter pylori (H. pylori). It colonizes in the stomach and is associated with severe gastric pathologies including gastric cancer and peptic ulceration. The most virulent factor of H. pylori is the cytotoxin-associated gene A (CagA) that is injected into the host cell. CagA interacts with several host proteins and alters their function, thereby causing several diseases. The most well-known target of CagA is the tumor suppressor protein ASPP2. The subdomain I at the N-terminus of CagA interacts with the proline-rich motif of ASPP2. Here, in this study, we carried out alanine scanning mutagenesis and an extensive molecular dynamics simulation summing up to 3.8 μs to find out hot spot residues and discovered some new protein-protein interaction (PPI)-modulating molecules. Our findings are in line with previous biochemical studies and further suggested new residues that are crucial for binding. The alanine scanning showed that mutation of Y207 and T211 residues to alanine decreased the binding affinity. Likewise, dynamics simulation and molecular mechanics with generalized Born surface area (MMGBSA) analysis also showed the importance of these two residues at the interface. A four-feature pharmacophore model was developed based on these two residues, and top 10 molecules were filtered from ZINC, NCI, and ChEMBL databases. The good binding affinity of the CHEMBL17319 and CHEMBL1183979 molecules shows the reliability of our adopted protocol for binding hot spot residues. We believe that our study provides a new insight for using CagA as the therapeutic target for gastric cancer treatment and provides a platform for a future experimental study.

Rapid characterization of chemical markers for discrimination of Moutan Cortex and its processed products by direct injection-based mass spectrometry profiling and metabolomic method

BackgroundProcessing of herbal medicines is a characteristic pharmaceutical technique in Traditional Chinese Medicine, which can reduce toxicity and side effect, improve the flavor and efficacy, and even change the pharmacological action entirely. It is significant and crucial to perform a method to find chemical markers for differentiating herbal medicines in different processed degrees. PurposeThe aim of this study was to perform a rapid and reasonable method to discriminate Moutan Cortex and its processed products, and to reveal the characteristics of chemical components depend on chemical markers. MethodsThirty batches of Moutan Cortex and its processed products, including 11 batches of Raw Moutan Cortex (RMC), 9 batches of Moutan Cortex Tostus (MCT) and 10 batches of Moutan Cortex Carbonisatus (MCC), were directly injected in electrospray ionization quadrupole time-of-flight mass spectrometry (ESI-QTOF MS) for rapid analysis in positive and negative mode. Without chromatographic separation, each run was completed within 3 min. The raw MS data were automatically extracted by background deduction and molecular feature (MF) extraction algorithm. In negative mode, a total of 452 MFs were obtained and then pretreated by data filtration and differential analysis. After that, the filtered 85 MFs were treated by principal component analysis (PCA) to reduce the dimensions. Subsequently, a partial least squares discrimination analysis (PLS-DA) model was constructed for differentiation and chemical markers detection of Moutan Cortex in different processed degrees. The positive mode data were treated as same as those in negative mode. ResultsRMC, MCT and MCC were successfully classified. Moreover, 14 and 3 chemical markers from negative and positive mode respectively, were screened by the combination of their relative peak areas and the parameter variable importance in the projection (VIP) values in PLS-DA model. The content changes of these chemical markers were employed in order to illustrate chemical changes of Moutan Cortex after processed. ConclusionThese results showed that the proposed method which combined non-targeted metabolomics analysis with multivariate statistics analysis is reasonable and effective. It could not only be applied to discriminate herbal medicines and their processing products, but also to reveal the characteristics of chemical components during processing.

Metabolomics Study of Cultivated Bulbus Fritillariae Cirrhosae at Different Growth Stages using UHPLC-QTOF-MS Coupled with Multivariate Data Analysis.

Bulbus fritillariae cirrhosae (known as Chuān bèi mǔ in China, BFC) contain fritillaria steroidal alkaloids as the bioactive ingredients and are widely used as traditional Chinese medicine for the treatment of cough and phlegm. Due to limited wild resources, the cultivated species are becoming predominantly used in Chinese traditional medicine markets. To assess the impact of different growth stages on the alkaloids of cultivated BFC and establish a reference for quality control and guidance for appropriate harvesting practices. The ultra-high performance liquid chromatography quadrupole time-of-flight mass spectrometry (UHPLC-QTOF-MS) metabolomic strategy was applied to determine potential chemical markers for the discrimination and quality control of cultivated BFC in different growth stages. The molecular feature extraction and multivariate statistical analysis were applied to alkaloid extraction and full metabolomic profiling of cultivated BFC for classification and marker compound characterisation. This approach allowed the establishment of a fast and efficient comparative multivariate analysis of the metabolite composition of 42 samples covering growth of cultivated BFC ranging in age from one to seven years old. Four alkaloid compounds were identified in cultivated BFC based on accurate mass, retention time, and MS/MS fragments. These compounds may be used as potential chemical markers for the classification and discrimination of cultivated BFC samples indifferent growth stages. The proposed analytical method in combination with multivariate statistical analysis comprised a useful and powerful strategy to explore the chemical ingredients and transforming mechanisms of cultivated BFC and for quality evaluation and control. Copyright © 2018 John Wiley & Sons, Ltd.