Accurate transcriptome-wide identification and quantification of alternative polyadenylation from RNA-seq data with APAIQ.

Alternative polyadenylation (APA) enables a gene to generate multiple transcripts with different 3' ends, which is dynamic across different cell types or conditions. Many computational methods have been developed to characterize sample-specific APA using the corresponding RNA-seq data, but suffered from high error rate on both polyadenylation site (PAS) identification and quantification of PAS usage (PAU), and bias toward 3' untranslated regions. Here we developed a tool for APA identification and quantification (APAIQ) from RNA-seq data, which can accurately identify PAS and quantify PAU in a transcriptome-wide manner. Using 3' end-seq data as the benchmark, we showed that APAIQ outperforms current methods on PAS identification and PAU quantification, including DaPars2, Aptardi, mountainClimber, SANPolyA, and QAPA. Finally, applying APAIQ on 421 RNA-seq samples from liver cancer patients, we identified >540 tumor-associated APA events and experimentally validated two intronic polyadenylation candidates, demonstrating its capacity to unveil cancer-related APA with a large-scale RNA-seq data set.

msRepDB: a comprehensive repetitive sequence database of over 80 000 species.

Repeats are prevalent in the genomes of all bacteria, plants and animals, and they cover nearly half of the Human genome, which play indispensable roles in the evolution, inheritance, variation and genomic instability, and serve as substrates for chromosomal rearrangements that include disease-causing deletions, inversions, and translocations. Comprehensive identification, classification and annotation of repeats in genomes can provide accurate and targeted solutions towards understanding and diagnosis of complex diseases, optimization of plant properties and development of new drugs. RepBase and Dfam are two most frequently used repeat databases, but they are not sufficiently complete. Due to the lack of a comprehensive repeat database of multiple species, the current research in this field is far from being satisfactory. LongRepMarker is a new framework developed recently by our group for comprehensive identification of genomic repeats. We here propose msRepDB based on LongRepMarker, which is currently the most comprehensive multi-species repeat database, covering >80 000 species. Comprehensive evaluations show that msRepDB contains more species, and more complete repeats and families than RepBase and Dfam databases. (https://msrepdb.cbrc.kaust.edu.sa/pages/msRepDB/index.html).

A sensitive repeat identification framework based on short and long reads.

Numerous studies have shown that repetitive regions in genomes play indispensable roles in the evolution, inheritance and variation of living organisms. However, most existing methods cannot achieve satisfactory performance on identifying repeats in terms of both accuracy and size, since NGS reads are too short to identify long repeats whereas SMS (Single Molecule Sequencing) long reads are with high error rates. In this study, we present a novel identification framework, LongRepMarker, based on the global de novo assembly and k-mer based multiple sequence alignment for precisely marking long repeats in genomes. The major characteristics of LongRepMarker are as follows: (i) by introducing barcode linked reads and SMS long reads to assist the assembly of all short paired-end reads, it can identify the repeats to a greater extent; (ii) by finding the overlap sequences between assemblies or chomosomes, it locates the repeats faster and more accurately; (iii) by using the multi-alignment unique k-mers rather than the high frequency k-mers to identify repeats in overlap sequences, it can obtain the repeats more comprehensively and stably; (iv) by applying the parallel alignment model based on the multi-alignment unique k-mers, the efficiency of data processing can be greatly optimized and (v) by taking the corresponding identification strategies, structural variations that occur between repeats can be identified. Comprehensive experimental results show that LongRepMarker can achieve more satisfactory results than the existing de novo detection methods (https://github.com/BioinformaticsCSU/LongRepMarker).

The Euscaphis japonica genome and the evolution of malvids.

Malvids is one of the largest clades of rosids, includes 58 families and exhibits remarkable morphological and ecological diversity. Here, we report a high-quality chromosome-level genome assembly for Euscaphis japonica, an early-diverging species within malvids. Genome-based phylogenetic analysis suggests that the unstable phylogenetic position of E. japonica may result from incomplete lineage sorting and hybridization event during the diversification of the ancestral population of malvids. Euscaphis japonica experienced two polyploidization events: the ancient whole genome triplication event shared with most eudicots (commonly known as the γ event) and a more recent whole genome duplication event, unique to E. japonica. By resequencing 101 samples from 11 populations, we speculate that the temperature has led to the differentiation of the evergreen and deciduous of E. japonica and the completely different population histories of these two groups. In total, 1012 candidate positively selected genes in the evergreen were detected, some of which are involved in flower and fruit development. We found that reddening and dehiscence of the E. japonica pericarp and long fruit-hanging time promoted the reproduction of E. japonica populations, and revealed the expression patterns of genes related to fruit reddening, dehiscence and abscission. The key genes involved in pentacyclic triterpene synthesis in E. japonica were identified, and different expression patterns of these genes may contribute to pentacyclic triterpene diversification. Our work sheds light on the evolution of E. japonica and malvids, particularly on the diversification of E. japonica and the genetic basis for their fruit dehiscence and abscission.

MultiNanopolish: refined grouping method for reducing redundant calculations in Nanopolish.

MOTIVATION: Compared with the second-generation sequencing technologies, the third-generation sequencing technologies allows us to obtain longer reads (average ∼10 kbps, maximum 900 kbps), but brings a higher error rate (∼15% error rate). Nanopolish is a variant and methylation detection tool based on hidden Markov model, which uses Oxford Nanopore sequencing data for signal-level analysis. Nanopolish can greatly improve the accuracy of assembly, whereas it is limited by long running time since most executive parts of Nanopolish is a serial and computationally expensive process. RESULTS: In this paper, we present an effective polishing tool, Multithreading Nanopolish (MultiNanopolish), which decomposes the whole process of iterative calculation in Nanopolish into small independent calculation tasks, making it possible to run this process in the parallel mode. Experimental results show that MultiNanopolish reduces running time by 50% with read-uncorrected assembler (Miniasm) and 20% with read-corrected assembler (Canu and Flye) based on 40 threads mode compared to the original Nanopolish. AVAILABILITY AND IMPLEMENTATION: MultiNanopolish is available at GitHub: https://github.com/BioinformaticsCSU/MultiNanopolish. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Large amplitude vibration of a cantilever actuated by a high-frequency pulsed laser.

Laser excitation based on the thermoelastic principle is effective for micro-scale actuation, enabled energy conversion from optical to mechanical. The major advantages lie in non-contact actuation, easy miniaturization, and integration. To avoid surface damage, the laser power per unit is limited, leading to several micrometers of the vibration. In this study, a pure nickel millimeter-sized cantilever is successfully actuated at a low-frequency resonance (around Hz) via a nanosecond pulsed laser. By modal interaction, the energy is transferred from a low-intensity, high-frequency (around kHz) excitation to a low-frequency response with millimeter amplitude. The stable low-frequency resonance of the cantilever was maintained by changing the laser pulse parameters and the illumination locations. We also present a method to control the vibration of the cantilever using a modulated wave (MW: the laser wave modulated by a rectangular wave). The cantilever's amplitude can be efficiently adjusted by changing the laser power or duty cycle of the MW. The resonance frequency of the cantilever also can be altered by optimizing the geometries to meet various actuation requirements. This study enables large actuation (up to tens of millimeters) by laser excitation, facilitating applications in precision manipulation, microfluidic mixing, lab-on-a-chip device, and other related micro actuation devices.

RepAHR: an improved approach for de novo repeat identification by assembly of the high-frequency reads.

BACKGROUND: Repetitive sequences account for a large proportion of eukaryotes genomes. Identification of repetitive sequences plays a significant role in many applications, such as structural variation detection and genome assembly. Many existing de novo repeat identification pipelines or tools make use of assembly of the high-frequency k-mers to obtain repeats. However, a certain degree of sequence coverage is required for assemblers to get the desired assemblies. On the other hand, assemblers cut the reads into shorter k-mers for assembly, which may destroy the structure of the repetitive regions. For the above reasons, it is difficult to obtain complete and accurate repetitive regions in the genome by using existing tools. RESULTS: In this study, we present a new method called RepAHR for de novo repeat identification by assembly of the high-frequency reads. Firstly, RepAHR scans next-generation sequencing (NGS) reads to find the high-frequency k-mers. Secondly, RepAHR filters the high-frequency reads from whole NGS reads according to certain rules based on the high-frequency k-mer. Finally, the high-frequency reads are assembled to generate repeats by using SPAdes, which is considered as an outstanding genome assembler with NGS sequences. CONLUSIONS: We test RepAHR on five data sets, and the experimental results show that RepAHR outperforms RepARK and REPdenovo for detecting repeats in terms of N50, reference alignment ratio, coverage ratio of reference, mask ratio of Repbase and some other metrics.

An efficient classification algorithm for NGS data based on text similarity.

With the advancement of high-throughput sequencing technologies, the amount of available sequencing data is growing at a pace that has now begun to greatly challenge the data processing and storage capacities of modern computer systems. Removing redundancy from such data by clustering could be crucial for reducing memory, disk space and running time consumption. In addition, it also has good performance on reducing dataset noise in some analysis applications. In this study, we propose a high-performance short sequence classification algorithm (HSC) for next generation sequencing (NGS) data based on efficient hash function and text similarity. First, HSC converts all reads into k-mers, then it forms a unique k-mer set by merging the duplicated and reverse complementary elements. Second, all unique k-mers are stored in a hash table, where the k-mer string is stored in the key field, and the ID of the reads containing the k-mer are stored in the value field. Third, each hash unit is transformed into a short text consisting of reads. Fourth, texts that satisfy the similarity threshold are combined into a long text, the merge operation is executed iteratively until there is no text that satisfies the merge condition. Finally, the long text is transformed into a cluster consisting of reads. We tested HSC using five real datasets. The experimental results showed that HSC cluster 100 million short reads within 2 hours, and it has excellent performance in reducing memory consumption. Compared to existing methods, HSC is much faster than other tools, it can easily handle tens of millions of sequences. In addition, when HSC is used as a preprocessing tool to produce assembly data, the memory and time consumption of the assembler is greatly reduced. It can help the assembler to achieve better assemblies in terms of N50, NA50 and genome fraction.

Genetic association of the gut microbiota with epigenetic clocks mediated by inflammatory cytokines: a Mendelian randomization analysis.

Background: A new aging biomarker epigenetic clock has been developed. There exists a close link between aging and gut microbiota, which may be mediated by inflammatory cytokines. However, the relationship between the epigenetic clock, gut microbiota, and the mediating substances is unclear. Methods: Two large genome-wide association meta-analyses were analyzed by two-sample Mendelian randomization. The results between gut microbiota and epigenetic clock were investigated using the four methods (Inverse variance weighted, MR-Egger, weighted median, MR-PRESSO). Genetic correlation was measured by Linked disequilibrium score regression (LDSC). The correctness of the study direction was checked by the Steiger test. Cochran's Q statistic and MR-Egger intercept were used as sensitivity analyses of the study. The two-step method was used to examine the mediating role of inflammatory cytokines. We use the Benjamini-Hochberg correction method to correct the P value. Results: After FDR correction, multiple bacterial genera were significantly or suggestively associated with four epigenetic clocks (GrimAge, HannumAge, IEAA, PhenoAge). And we detected several inflammatory factors acting as mediators of gut microbiota and epigenetic clocks. Conclusion: This study provides genetic evidence for a positive and negative link between gut microbiota and aging risk. We hope that by elucidating the genetic relationship and potential mechanisms between aging and gut microbiota, we will provide new avenues for continuing aging-related research and treatment.

Expression of Toll-like receptors in the cerebellum during pathogenesis of prion disease.

Prion diseases, such as scrapie, entail the accumulation of disease-specific prion protein (PrPSc) within the brain. Toll-like receptors (TLRs) are crucial components of the pattern recognition system. They recognize pathogen-associated molecular patterns (PAMPs) and play a central role in orchestrating host innate immune responses. The expression levels of Toll-like receptors (TLRs) in the central nervous system (CNS) were not well-defined. To establish a model of prion diseases in BALB/C mice, the 22L strain was employed. The features of the 22L strain were analyzed, and the cerebellum exhibited severe pathological changes. TLR1-13 levels in the cerebellum were measured using quantitative polymerase chain reaction (qPCR) at time points of 60, 90, 120, and the final end point (145 days post-infection). During the pathogenesis, the expression levels of Toll-like receptors (TLRs) 1, 2, 7, 8, and 9 increased in a time-dependent manner. This trend mirrored the expression patterns of PrPSc (the pathological isoform of the prion protein) and glial fibrillary acidic protein. Notably, at the end point, TLR1-13 levels were significantly elevated. Protein level of TLR7 and TLR9 showed increasing at the end point of the 22L-infected mice. A deeper understanding of the increased Toll-like receptors (TLRs) in prion diseases could shed light on their role in initiating immune responses at various stages during pathogenesis. This insight is particularly relevant when considering TLRs as potential therapeutic targets for prion diseases.

Pre-trained multimodal large language model enhances dermatological diagnosis using SkinGPT-4.

Large language models (LLMs) are seen to have tremendous potential in advancing medical diagnosis recently, particularly in dermatological diagnosis, which is a very important task as skin and subcutaneous diseases rank high among the leading contributors to the global burden of nonfatal diseases. Here we present SkinGPT-4, which is an interactive dermatology diagnostic system based on multimodal large language models. We have aligned a pre-trained vision transformer with an LLM named Llama-2-13b-chat by collecting an extensive collection of skin disease images (comprising 52,929 publicly available and proprietary images) along with clinical concepts and doctors' notes, and designing a two-step training strategy. We have quantitatively evaluated SkinGPT-4 on 150 real-life cases with board-certified dermatologists. With SkinGPT-4, users could upload their own skin photos for diagnosis, and the system could autonomously evaluate the images, identify the characteristics and categories of the skin conditions, perform in-depth analysis, and provide interactive treatment recommendations.

Stereotactic radiotherapy vs whole brain radiation therapy in EGFR mutated NSCLC: Results & reflections from the prematurely closed phase III HYBRID trial.

BACKGROUND: All known randomized trials of stereotactic radiotherapy (SRT) versus whole brain radiotherapy (WBRT) for brain metastases (BMs) comprise mixed histologies. The phase III HYBRID trial (NCT02882984) attempted to evaluate the non-inferiority of SRT vs. WBRT specifically for EGFR-mutated non-small cell lung cancer (EGFRm NSCLC) BMs. METHODS: Inclusion criteria were ≤ 5 BMs (any size) from treatment-naïve EGFRm NSCLC. All patients started a first-generation tyrosine kinase inhibitor on the first day of WBRT (37.5 Gy/15 fractions) or SRT (25-40 Gy/5 fractions per tumor volume). The primary endpoint was 18-month intracranial progression-free survival (iPFS; intention-to-treat). RESULTS: The trial commenced in June 2015 and was closed in April 2021 after screening 208 patients but enrolling 85 (n = 41 WBRT, n = 44 SRT; median follow-up 31 and 36 months, respectively). Respectively, 9.5 % vs. 10.2 % of patients experienced intracranial progression at 18 months, and the median iPFS was 21.4 vs. 22.3 months (p > 0.05 for all). The SRT arm experienced higher overall survival and cognitive preservation (p < 0.05 for all). The most notable reason for low enrollment was patients not wishing to risk neurocognitive decline from WBRT. CONCLUSIONS: Although this phase III trial was underpowered, there was no evidence that SRT yielded outcome detriments compared to WBRT for EGFRm NSCLC BMs. Lessons from prematurely closed trials are valuable, as they often provide important experiential perspectives for investigators designing/executing future trials. In the current era, randomized trials involving WBRT without cognitive sparing measures may be at high risk of underaccrual; trial investigators are encouraged to carefully consider our experience when attempting to design such trials. However, trials of molecular-/biologically-stratified patients are highly recommended as the notion of "individualized medicine/oncology" continues to expand.

LCAT: an isoform-sensitive error correction for transcriptome sequencing long reads.

As the carrier of genetic information, RNA carries the information from genes to proteins. Transcriptome sequencing technology is an important way to obtain transcriptome sequences, and it is also the basis for transcriptome research. With the development of third-generation sequencing, long reads can cover full-length transcripts and reflect the composition of different isoforms. However, the high error rate of third-generation sequencing affects the accuracy of long reads and downstream analysis. The current error correction methods seldom consider the existence of different isoforms in RNA, which makes the diversity of isoforms a serious loss. Here, we introduce LCAT (long-read error correction algorithm for transcriptome sequencing data), a wrapper algorithm of MECAT, to reduce the loss of isoform diversity while keeping MECAT's error correction performance. The experimental results show that LCAT can not only improve the quality of transcriptome sequencing long reads but also retain the diversity of isoforms.

AB-Gen: Antibody Library Design with Generative Pre-trained Transformer and Deep Reinforcement Learning.

Antibody leads must fulfill multiple desirable properties to be clinical candidates. Primarily due to the low throughput in the experimental procedure, the need for such multi-property optimization causes the bottleneck in preclinical antibody discovery and development, because addressing one issue usually causes another. We developed a reinforcement learning (RL) method, named AB-Gen, for antibody library design using a generative pre-trained transformer (GPT) as the policy network of the RL agent. We showed that this model can learn the antibody space of heavy chain complementarity determining region 3 (CDRH3) and generate sequences with similar property distributions. Besides, when using human epidermal growth factor receptor-2 (HER2) as the target, the agent model of AB-Gen was able to generate novel CDRH3 sequences that fulfill multi-property constraints. Totally, 509 generated sequences were able to pass all property filters, and three highly conserved residues were identified. The importance of these residues was further demonstrated by molecular dynamics simulations, consolidating that the agent model was capable of grasping important information in this complex optimization task. Overall, the AB-Gen method is able to design novel antibody sequences with an improved success rate than the traditional propose-then-filter approach. It has the potential to be used in practical antibody design, thus empowering the antibody discovery and development process. The source code of AB-Gen is freely available at Zenodo (https://doi.org/10.5281/zenodo.7657016) and BioCode (https://ngdc.cncb.ac.cn/biocode/tools/BT007341).

Repetitive DNA sequence detection and its role in the human genome.

Repetitive DNA sequences playing critical roles in driving evolution, inducing variation, and regulating gene expression. In this review, we summarized the definition, arrangement, and structural characteristics of repeats. Besides, we introduced diverse biological functions of repeats and reviewed existing methods for automatic repeat detection, classification, and masking. Finally, we analyzed the type, structure, and regulation of repeats in the human genome and their role in the induction of complex diseases. We believe that this review will facilitate a comprehensive understanding of repeats and provide guidance for repeat annotation and in-depth exploration of its association with human diseases.

A unified method to revoke the private data of patients in intelligent healthcare with audit to forget.

Revoking personal private data is one of the basic human rights. However, such right is often overlooked or infringed upon due to the increasing collection and use of patient data for model training. In order to secure patients' right to be forgotten, we proposed a solution by using auditing to guide the forgetting process, where auditing means determining whether a dataset has been used to train the model and forgetting requires the information of a query dataset to be forgotten from the target model. We unified these two tasks by introducing an approach called knowledge purification. To implement our solution, we developed an audit to forget software (AFS), which is able to evaluate and revoke patients' private data from pre-trained deep learning models. Here, we show the usability of AFS and its application potential in real-world intelligent healthcare to enhance privacy protection and data revocation rights.

Soil nutrients, enzyme activities, and bacterial communities in varied plant communities in karst rocky desertification regions in Wushan County, Southwest China.

Vegetation restoration has become a common practice in karst rocky desertification (KRD) areas of southwestern China. The bacteria, which have made a connection between soil and plants, have been an important role in regulating the succession and restoration of karst vegetation. However, it is still unclear how soil bacterial communities and soil properties respond to natural vegetation restoration processes in karst areas. To address this gap, we investigated the soil nutrients, enzyme activity, and soil bacterial community among various plant communities, including farmland (FL), land with herbs only (SSI), herb-and-shrub land (SSII), woody thickets (SSIII), coniferous forest (SSIV), coniferous and broad-leaved mixed forest (SSV), and evergreen broad-leaved forest (SSVI). Our results found that SSII had the highest levels of soil organic matter, total nitrogen, available phosphorus, available nitrogen, sucrase, and ß-glucosidase among all the plant communities. These results indicated that herb-and-shrub land have contributed to the rapid restoration of vegetation in KRD regions. FL exhibited the lowest levels of soil nutrients and enzyme activities, while showing the highest bacterial richness and diversity among all the plant communities. This suggested that appropriate human intervention can increase bacterial diversity and richness in the area. The predominant bacterial phylum also varied among the different plant communities, with Actinobacteria being the most abundant in SSI, SSII, SSIII, and SSIV, while Proteobacteria were the most abundant in SSV and SSVI. Furthermore, PCoA analysis demonstrated significant changes in the soil bacterial community structure, with SSI, SSII, SSIII, and SSIV had shared similar structures, while SSV and SSVI had comparable structures. As for soil characteristics, total phosphorus (TP) and total potassium (TK) were the primary factors affecting the soil bacterial community. SSV and SSVI had the most complex bacterial networks and were more stable than other groups. The genera Ktedonobacter, norank_f_Anaerolineaceae, and Vicinamibacter had the highest betweenness centrality scores and were identified as keystone genera in the co-occurrence network in KRD areas. In summary, our results have demonstrated that herb-and-shrub can promote community succession and increase soil nutrient levels in KRD regions.

A comprehensive benchmarking with practical guidelines for cellular deconvolution of spatial transcriptomics.

Spatial transcriptomics technologies are used to profile transcriptomes while preserving spatial information, which enables high-resolution characterization of transcriptional patterns and reconstruction of tissue architecture. Due to the existence of low-resolution spots in recent spatial transcriptomics technologies, uncovering cellular heterogeneity is crucial for disentangling the spatial patterns of cell types, and many related methods have been proposed. Here, we benchmark 18 existing methods resolving a cellular deconvolution task with 50 real-world and simulated datasets by evaluating the accuracy, robustness, and usability of the methods. We compare these methods comprehensively using different metrics, resolutions, spatial transcriptomics technologies, spot numbers, and gene numbers. In terms of performance, CARD, Cell2location, and Tangram are the best methods for conducting the cellular deconvolution task. To refine our comparative results, we provide decision-tree-style guidelines and recommendations for method selection and their additional features, which will help users easily choose the best method for fulfilling their concerns.

ZnO@Carbon Dot Nanoparticles Stimulating the Antibacterial Activity of Polyvinylidene Fluoride-Hexafluoropropylene with a Higher Electroactive Phase for Multifunctional Devices.

To further advance the application of flexible piezoelectric materials in wearable/implantable devices and robot electronic skin, it is necessary to endow them with a new function of antibacterial properties and with higher piezoelectric performance. Introducing a specially designated nanomaterial based on the nanocomposite effect is a feasible strategy to improve material properties and achieve multifunctionalization of composites. In this paper, carbon dots (CDs) were sensitized onto the surface of ZnO to form ZnO@CDs nanoparticles, which were then incorporated into polyvinylidene fluoride-hexafluoropropylene (PVDF-HFP) to obtain a multifunctional composite. On the one hand, the antibacterial property of ZnO was improved because CDs had good optical absorption of visible light and their surface functional groups were favorable for electrostatic adsorption with bacteria. Therefore, ZnO@CDs endowed the composite with an outstanding antibacterial rate of 69.1% for Staphylococcus aureus. On the other hand, CDs played a bridging role between ZnO and PVDF-HFP, reducing the negative effect of ZnO aggregation and interface incompatibility with PVDF-HFP. As a result, ZnO@CDs induced ß-phase formation of 80.4% in PVDF-HFP with a d33 value of 33.8 pC N-1. The multifunctional device exhibited excellent piezoelectric and antibacterial performance in the application of energy harvesters and self-powered pressure sensors.

Optimization of binding affinities in chemical space with generative pre-trained transformer and deep reinforcement learning.

Background: The key challenge in drug discovery is to discover novel compounds with desirable properties. Among the properties, binding affinity to a target is one of the prerequisites and usually evaluated by molecular docking or quantitative structure activity relationship (QSAR) models. Methods: In this study, we developed SGPT-RL, which uses a generative pre-trained transformer (GPT) as the policy network of the reinforcement learning (RL) agent to optimize the binding affinity to a target. SGPT-RL was evaluated on the Moses distribution learning benchmark and two goal-directed generation tasks, with Dopamine Receptor D2 (DRD2) and Angiotensin-Converting Enzyme 2 (ACE2) as the targets. Both QSAR model and molecular docking were implemented as the optimization goals in the tasks. The popular Reinvent method was used as the baseline for comparison. Results: The results on the Moses benchmark showed that SGPT-RL learned good property distributions and generated molecules with high validity and novelty. On the two goal-directed generation tasks, both SGPT-RL and Reinvent were able to generate valid molecules with improved target scores. The SGPT-RL method achieved better results than Reinvent on the ACE2 task, where molecular docking was used as the optimization goal. Further analysis shows that SGPT-RL learned conserved scaffold patterns during exploration. Conclusions: The superior performance of SGPT-RL in the ACE2 task indicates that it can be applied to the virtual screening process where molecular docking is widely used as the criteria. Besides, the scaffold patterns learned by SGPT-RL during the exploration process can assist chemists to better design and discover novel lead candidates.