Sequencing Machines Research Articles

The influence of preformed metal crowns versus zirconia crowns on the diagnostic quality of magnetic resonance images.

The purpose of this study was to investigate and compare in vitro the influence of preformed metal crowns (PMC) versus zirconia crowns (ZC) on the extent of artifact formation and the diagnostic quality of magnetic resonance imaging (MRI). Standard crown tooth preparations were performed on Typodont models. PMC and ZC were cemented on all primary molar teeth (n = 8) of the respective models. MRI scans were taken for the Typodont alone (control), PMC, and ZC, using a 3T MRI machine and common MRI sequences. The diameter of image distortion was measured around the dental materials. Data was analyzed using a mixed model analysis. The average artifact size for PMC was 54.9mm (range 19.6-110.7mm) and 12.8mm (range 5.5-20.7mm) for ZC. The artifact size was associated with the MRI pulse sequence and dental material, with PMC showing larger artifact diameters than ZC for all MRI sequences (p < 0.001). The control had sensitive sequences for AxT2* and DWI sequences with MRI artifacts of 11.4mm and 10.7mm, respectively. It is important that dentists are aware of the potential of dental materials to cause unfavourable MRI artifacts. When areas of the brain requiring MRI preclude the use of PMC, ZC may be considered as an alternative to extractions. Further clinical studies are required to evaluate the impact of dental materials on the image quality of MR scans in vivo.

Detection and molecular characterization of carbapenem-resistant gram-negative bacterial isolates.

Antimicrobial-resistant bacteria (ARB) are responsible for increased mortality and morbidity. Therefore, this study focuses on evaluating traditional and molecular diagnostic tools of carbapenem-resistant gram-negative bacteria (CRGNB). In order to achieve this, 94 samples, from different patients' specimens, and surrounding environment, were collected from intensive care units (ICUs) at Ain Shams University Specialized Hospital and the National Cancer Institute, Cairo, Egypt. The swabs were cultured on appropriate media, including Chromogenic medium (HiCrome KPC Agar Base "HIMEDI AM1831"), and MacConkey-10µg imipenem disc resulting in 136 isolates with different culture characteristics. Next, the selected isolates were subjected to VITEK 2 machine and 16SrRNA (16S ribosomal RNA) sequencing. The sensitivity of HiCROME KPC agar for CRGNB detection was 99.3% and 94.7%, in reference to the MacConkey-disc and VITEK-2 methods, respectively. The HiCrome KPC agar assumptions for bacterial identification were not as consistent as those of VITEK 2 (with only 47.4% agreement) and 16SrRNA gene sequencing analysis. The approaches discussed in this study facilitate providing rapid diagnosis and treatment of CRGNB, which helps increase survival rates. HiCrome KPC agar is considered a relatively accurate and easy method that can be used in any laboratory. In addition, the selected strains were deposited in the gene bank with the accession numbers OR553657, OR553658, and OR553659. It is noteworthy that Genus Acinetobacter is the major CRGNB isolated from the patients and environmental surfaces in the hospitals. This highlights the importance of proper environmental and terminal cleaning procedures in healthcare facilities and applying control measures to ensure infection prevention.

A hybrid niching memetic algorithm for multi-modal optimization of double row layout problem

Double row layout problem (DRLP) involves identifying the exact locations of machines participating in a production task on two rows. There are typically multiple layouts with approximately optimal material handling cost for a DRLP. These layouts often exhibit significantly different layout configurations. Identifying multiple global or local optimal layouts can provide layout designers with a wide range of options, which is of great significance for enhancing the maintainability, scalability, and customisability of the facility. However, most existing studies on DRLPs typically focus on designing a single optimal layout. In this paper, we study a multi-modal optimization of double row layout problem (MDRLP). A hybrid approach combing a fast niching memetic algorithm and linear programming (FNMA-LP) is proposed for MDRLP to locate multiple global or local optimal layouts with a similar quality. First, a fast niching memetic algorithm is developed to find a set of approximate optimal machine sequences. Then, LP is employed to optimise the exact locations of machines for each machine sequence. To evaluate the performance of the proposed algorithm, FNMA-LP is compared against three popular multi-modal algorithms and a state-of-the-art single-modal algorithm developed for DRLP. Experiments show that our approach outperforms competing approaches on almost all problem instances.

Machine learning based analysis of single-cell data reveals evidence of subject-specific single-cell gene expression profiles in acute myeloid leukaemia patients and healthy controls

Acute Myeloid Leukaemia (AML) is characterized by uncontrolled growth of immature myeloid cells, disrupting normal blood production. Treatment typically involves chemotherapy, targeted therapy, and stem cell transplantation but many patients develop chemoresistance, leading to poor outcomes due to the disease's high heterogeneity. In this study, we used publicly available single-cell RNA sequencing data and machine learning to classify AML patients and healthy, monocytes, dendritic and progenitor cells population. We found that gene expression profiles of AML patients and healthy controls can be classified at the individual level with high accuracy (>70 %) when using progenitor cells, suggesting the existence of subject-specific single cell transcriptomics profiles. The analysis also revealed molecular determinants of patient heterogeneity (e.g. TPSD1, CT45A1, and GABRA4) which could support new strategies for patient stratification and personalized treatment in leukaemia.

Gut microbiota, serum metabolites, and lipids related to blood glucose control and type 1 diabetes.

The composition and function of gut microbiota, lipids, and metabolites in patients with type 1 diabetes (T1D) or its association with glycemic control remains unknown. We aimed to use multi-omics sequencing technology and machine learning (ML) approaches to investigate potential function and relationships among the gut microbiota, lipids, and metabolites in T1D patients at varied glycemic levels. We conducted a multi-omics analysis of the gut microbiome from fecal samples, metabolites, and lipids obtained from serum samples, collected from a cohort of 72 T1D patients. The patients were divided into two groups based on their hemoglobin A1c (HbA1c) levels. 16S rRNA sequencing, and metabolomics methods were applied to analyze changes in composition and function of gut microbiota, metabolites, and lipids. The linear discriminant analysis, Shapley additive explanations (SHAP) algorithm, and ML algorithms revealed the enrichment of Bacteroides_nordii, Bacteroides_cellulosilyticus in the glycemic control (GC) group, while Bacteroides_coprocola and Sutterella_wadsworthensis were enriched in the poor glycemic control (PGC) group. Several metabolic enrichment sets like fatty acid biosynthesis and glycerol phosphate shuttle metabolism were different between two groups. Bacteroides_nordii exhibited a negative association with D-fructose, a component involved in the starch and sucrose metabolism pathway, as well as with monoglycerides (16:0) involved in the glycerolipid metabolism pathway. We identified distinct characteristics of gut microbiota, metabolites, and lipids in T1D patients exhibiting different levels of glycemic control. Through comprehensive analysis, microbiota (Bacteroides_nordii, Bacteroides_coprocola), metabolites (D-fructose), and lipids (Monoglycerides) may serve as potential mediators that communicated the interaction between the gut, circulatory systems, and glucose fluctuations in T1D patients.

Enhancing Arabidopsis thaliana ubiquitination site prediction through knowledge distillation and natural language processing

Protein ubiquitination is a critical post-translational modification (PTM) involved in diverse biological processes and plays a pivotal role in regulating physiological mechanisms and disease states. Despite various efforts to develop ubiquitination site prediction tools across species, these tools mainly rely on predefined sequence features and machine learning algorithms, with species-specific variations in ubiquitination patterns remaining poorly understood. This study introduces a novel approach for predicting Arabidopsis thaliana ubiquitination sites using a neural network model based on knowledge distillation and natural language processing (NLP) of protein sequences. Our framework employs a multi-species “Teacher model” to guide a more compact, species-specific “Student model”, with the “Teacher” generating pseudo-labels that enhance the “Student” learning and prediction robustness. Cross-validation results demonstrate that our model achieves superior performance, with an accuracy of 86.3 % and an area under the curve (AUC) of 0.926, while independent testing confirmed these results with an accuracy of 86.3 % and an AUC of 0.923. Comparative analysis with established predictors further highlights the model’s superiority, emphasizing the effectiveness of integrating knowledge distillation and NLP in ubiquitination prediction tasks. This study presents a promising and efficient approach for ubiquitination site prediction, offering valuable insights for researchers in related fields. The code and resources are available on GitHub: https://github.com/nuinvtnu/KD_ArapUbi.

Multi-omics features of immunogenic cell death in gastric cancer identified by combining single-cell sequencing analysis and machine learning

Gastric cancer (GC) is a prevalent malignancy with high mortality rates. Immunogenic cell death (ICD) is a unique form of programmed cell death that is closely linked to antitumor immunity and plays a critical role in modulating the tumor microenvironment (TME). Nevertheless, elucidating the precise effect of ICD on GC remains a challenging endeavour. ICD-related genes were identified in single-cell sequencing datasets and bulk transcriptome sequencing datasets via the AddModuleScore function, weighted gene co-expression network (WGCNA), and differential expression analysis. A robust signature associated with ICD was constructed using a machine learning computational framework incorporating 101 algorithms. Furthermore, multiomics analysis, including single-cell sequencing analysis, bulk transcriptomic analysis, and proteomics analysis, was conducted to verify the correlation of these hub genes with the immune microenvironment features of GC and with GC invasion and metastasis. We screened 59 genes associated with ICD and developed a robust ICD-related gene signature (ICDRS) via a machine learning computational framework that integrates 101 different algorithms. Furthermore, we identified five key hub genes (SMAP2, TNFAIP8, LBH, TXNIP, and PIK3IP1) from the ICDRS. Through single-cell analysis of GC tumor s, we confirmed the strong correlations of the hub genes with immune microenvironment features. Among these five genes, LBH exhibited the most significant associations with a poor prognosis and with the invasion and metastasis of GC. Finally, our findings were validated through immunohistochemical staining of a large clinical sample set, and the results further supported that LBH promotes GC cell invasion by activating the epithelial–mesenchymal transition (EMT) pathway.

Polymers for Disrupting Protein-Protein Interactions: Where Are We and Where Should We Be?

Protein-protein interactions (PPIs) are central to the cellular signaling and regulatory networks that underlie many physiological and pathophysiological processes. It is challenging to target PPIs using traditional small molecule or peptide-based approaches due to the frequent lack of well-defined binding pockets at the large and flat PPI interfaces. Synthetic polymers offer an opportunity to circumvent these challenges by providing unparalleled flexibility in tuning their physiochemical properties to achieve the desired binding properties. In this review, we summarize the current state of the field pertaining to polymer-protein interactions in solution, highlighting various polyelectrolyte systems, their tunable parameters, and their characterization. We provide an outlook on how these architectures can be improved by incorporating sequence control, foldability, and machine learning to mimic proteins at every structural level. Advances in these directions will enable the design of more specific protein-binding polymers and provide an effective strategy for targeting dynamic proteins, such as intrinsically disordered proteins.

Lightweight taxonomic profiling of long-read metagenomic datasets with Lemur and Magnet.

The advent of long-read sequencing of microbiomes necessitates the development of new taxonomic profilers tailored to long-read shotgun metagenomic datasets. Here, we introduce Lemur and Magnet, a pair of tools optimized for lightweight and accurate taxonomic profiling for long-read shotgun metagenomic datasets. Lemur is a marker-gene-based method that leverages an EM algorithm to reduce false positive calls while preserving true positives; Magnet is a whole-genome read mapping based method that provides detailed presence and absence calls for bacterial genomes. We demonstrate that Lemur and Magnet can run in minutes to hours on a laptop with 32 GB of RAM, even for large inputs, a crucial feature given the portability of long-read sequencing machines. Furthermore, the marker gene database used by Lemur is only 4 GB and contains information from over 300,000 RefSeq genomes. Lemur and Magnet are open-source and available at https://github.com/treangenlab/lemur and https://github.com/treangenlab/magnet.

Anomaly detection for multivariate time series in IoT using discrete wavelet decomposition and dual graph attention networks

Effective anomaly detection in multivariate time series data is critical to ensuring the security of Internet of Things (IoT) devices and systems. However, building a high precision and low false positive rate anomaly detection model for the complex and volatile IoT environment is a challenging task. This is often due to issues such as a lack of anomaly labeling, high data volatility, and the complexity of device mechanisms. Traditional machine learning algorithms and sequence models frequently fail to account for feature correlation and temporal dependency in anomaly detection. Although deep learning-based anomaly detection methods have progressed, there is still room for improvement in precision, recall, and generalization ability. In this paper, we propose an anomaly detection model called Meta-MWDG to address these issues. The model is based on a multi-scale discrete wavelet decomposition and a dual graph attention network, which can effectively extract feature correlation and temporal dependency in multivariate time series data. Additionally, model-agnostic meta-learning (MAML) is introduced to improve the model’s generalization performance, enabling it to perform well on new tasks even with a few samples. A gated recurrent unit (GRU) is combined with a multi-head self-attention network to output both prediction and reconstruction results in a joint optimization strategy, improving the precision of anomaly detection. Extensive experimental studies demonstrate that Meta-MWDG outperforms the state-of-the-art methods in anomaly detection.

Molecular precision medicine: Multi-omics-based stratification model for acute myeloid leukemia

Acute myeloid leukemia (AML), as the most common malignancy of the hematopoietic system, poses challenges in treatment efficacy, relapse, and drug resistance. In this study, we have utilized 151 RNA sequencing datasets, 194 DNA methylation datasets, and 200 somatic mutation datasets from the AML cohort in the TCGA database to develop a multi-omics stratification model. This model enables comparison of prognosis, clinical features, gene mutations, immune microenvironment and drug sensitivity across subgroups. External validation datasets have been sourced from the GEO database, which includes 562 mRNA datasets and 136 miRNA datasets from 984 adult AML patients. Through multi-omics-based stratification model, we classified 126 AML patients into 4 clusters (CS). CS4 had the best prognosis, with the youngest age, highest M3 subtype proportion, fewest copy number alterations, and common mutations in WT1, FLT3, and KIT genes. It showed sensitivity to HDAC inhibitors and BCL-2 inhibitors. Both the M3 subtype and CS4 were identified as independent protective factors for survival. Conversely, CS3 had the worst prognosis due to older age, high copy number alterations, and frequent mutations in RUNX1, DNMT3A, and TP53 genes. Additionally, it showed higher proportions of cytotoxic cells and Tregs, suggesting potential sensitivity to mTOR inhibitors. CS1 had a better prognosis than CS2, with more copy number alterations, while CS2 had higher monocyte proportions. CS1 showed good sensitivity to cytarabine, while CS2 was sensitive to RXR agonists. Both CS1 and CS2, which predominantly featured mutations in FLT3, NPM1, and DNMT3A genes, benefited from FLT3 inhibitors. Using the Kappa test, our stratification model underwent robust validation in the miRNA and mRNA external validation datasets. With advancements in sequencing technology and machine learning algorithms, AML is poised to transition towards multi-omics precision medicine in the future. We aspire for our study to offer new perspectives on multi-drug combination clinical trials and multi-targeted precision medicine for AML.

Lipid discovery enabled by sequence statistics and machine learning.

Bacterial membranes are complex and dynamic, arising from an array of evolutionary pressures. One enzyme that alters membrane compositions through covalent lipid modification is MprF. We recently identified that Streptococcus agalactiae MprF synthesizes lysyl-phosphatidylglycerol (Lys-PG) from anionic PG, and a novel cationic lipid, lysyl-glucosyl-diacylglycerol (Lys-Glc-DAG), from neutral glycolipid Glc-DAG. This unexpected result prompted us to investigate whether Lys-Glc-DAG occurs in other MprF-containing bacteria, and whether other novel MprF products exist. Here, we studied protein sequence features determining MprF substrate specificity. First, pairwise analyses identified several streptococ-cal MprFs synthesizing Lys-Glc-DAG. Second, a restricted Boltzmann machine-guided approach led us to discover an entirely new substrate for MprF in Enterococcus , diglucosyl-diacylglycerol (Glc 2 -DAG), and an expanded set of organisms that modify glycolipid substrates using MprF. Overall, we combined the wealth of available sequence data with machine learning to model evolutionary constraints on MprF sequences across the bacterial domain, thereby identifying a novel cationic lipid.

Quantitative drug susceptibility testing for Mycobacterium tuberculosis using unassembled sequencing data and machine learning.

There remains a clinical need for better approaches to rapid drug susceptibility testing in view of the increasing burden of multidrug resistant tuberculosis. Binary susceptibility phenotypes only capture changes in minimum inhibitory concentration when these cross the critical concentration, even though other changes may be clinically relevant. We developed a machine learning system to predict minimum inhibitory concentration from unassembled whole-genome sequencing data for 13 anti-tuberculosis drugs. We trained, validated and tested the system on 10,859 isolates from the CRyPTIC dataset. Essential agreement rates (predicted MIC within one doubling dilution of observed MIC) were above 92% for first-line drugs, 91% for fluoroquinolones and aminoglycosides, and 90% for new and repurposed drugs, albeit with a significant drop in performance for the very few phenotypically resistant isolates in the latter group. To further validate the model in the absence of external MIC datasets, we predicted MIC and converted values to binary for an external set of 15,239 isolates with binary phenotypes, and compare their performance against a previously validated mutation catalogue, the expected performance of existing molecular assays, and World Health Organization Target Product Profiles. The sensitivity of the model on the external dataset was greater than 90% for all drugs except ethionamide, clofazimine and linezolid. Specificity was greater than 95% for all drugs except ethambutol, ethionamide, bedaquiline, delamanid and clofazimine. The proposed system can provide quantitative susceptibility phenotyping to help guide antimicrobial therapy, although further data collection and validation are required before machine learning can be used clinically for all drugs.

Leveraging cfDNA fragmentomic features in a stacked ensemble model for early detection of esophageal squamous cell carcinoma

In this study, we develop a stacked ensemble model that utilizes cell-free DNA (cfDNA) fragmentomics for the early detection of esophageal squamous cell carcinoma (ESCC). This model incorporates four distinct fragmentomics features derived from whole-genome sequencing (WGS) and advanced machine learning algorithms for robust analysis. It is validated across both an independent validation cohort and an external cohort to ensure its generalizability and effectiveness. Notably, the model maintains its robustness in low-coverage sequencing environments, demonstrating its potentials in clinical settings with limited sequencing resources. With its remarkable sensitivity and specificity, this approach promises to significantly improve the early diagnosis and management of ESCC. This study represents a substantial step forward in the application of cfDNA fragmentomics in cancer diagnostics, emphasizing the need for further research to fully establish its clinical efficacy.

Exploring the Genotype-Phenotype Correlations in a Child with Inherited Seizure and Thrombocytopenia by Digenic Network Analysis.

Understanding the correlation between genotype and phenotype remains challenging for modern genetics. Digenic network analysis may provide useful models for understanding complex phenotypes that traditional Mendelian monogenic models cannot explain. Clinical data, whole exome sequencing data, in silico, and machine learning analysis were combined to construct a digenic network that may help unveil the complex genotype-phenotype correlations in a child presenting with inherited seizures and thrombocytopenia. The proband inherited a maternal heterozygous missense variant in SCN1A (NM_001165963.4:c.2722G>A) and a paternal heterozygous missense variant in MYH9 (NM_002473.6:c.3323A>C). In silico analysis showed that these two variants may be pathogenic for inherited seizures and thrombocytopenia in the proband. Moreover, focusing on 230 epilepsy-associated genes and 35 thrombopoiesis genes, variant call format data of the proband were analyzed using machine learning tools (VarCoPP 2.0) and Digenic Effect predictor. A digenic network was constructed, and SCN1A and MYH9 were found to be core genes in the network. Further analysis showed that MYH9 might be a modifier of SCN1A, and the variant in MYH9 might not only influence the severity of SCN1A-related seizure but also lead to thrombocytopenia in the bone marrow. In addition, another eight variants might also be co-factors that account for the proband's complex phenotypes. Our data show that as a supplement to the traditional Mendelian monogenic model, digenic network analysis may provide reasonable models for the explanation of complex genotype-phenotype correlations.

Machine learning based on metabolomics unveils neutrophil extracellular trap-related metabolic signatures in non-small cell lung cancer patients undergoing chemoimmunotherapy.

Non-small cell lung cancer (NSCLC) is the primary form of lung cancer, and the combination of chemotherapy with immunotherapy offers promising treatment options for patients suffering from this disease. However, the emergence of drug resistance significantly limits the effectiveness of these therapeutic strategies. Consequently, it is imperative to devise methods for accurately detecting and evaluating the efficacy of these treatments. To identify the metabolic signatures associated with neutrophil extracellular traps (NETs) and chemoimmunotherapy efficacy in NSCLC patients. In total, 159 NSCLC patients undergoing first-line chemoimmunotherapy were enrolled. We first investigated the characteristics influencing clinical efficacy. Circulating levels of NETs and cytokines were measured by commercial kits. Liquid chromatography tandem mass spectrometry quantified plasma metabolites, and differential metabolites were identified. Least absolute shrinkage and selection operator, support vector machine-recursive feature elimination, and random forest algorithms were employed. By using plasma metabolic profiles and machine learning algorithms, predictive metabolic signatures were established. First, the levels of circulating interleukin-8, neutrophil-to-lymphocyte ratio, and NETs were closely related to poor efficacy of first-line chemoimmunotherapy. Patients were classed into a low NET group or a high NET group. A total of 54 differential plasma metabolites were identified. These metabolites were primarily involved in arachidonic acid and purine metabolism. Three key metabolites were identified as crucial variables, including 8,9-epoxyeicosatrienoic acid, L-malate, and bis(monoacylglycerol)phosphate (18:1/16:0). Using metabolomic sequencing data and machine learning methods, key metabolic signatures were screened to predict NET level as well as chemoimmunotherapy efficacy. The identified metabolic signatures may effectively distinguish NET levels and predict clinical benefit from chemoimmunotherapy in NSCLC patients.

Deciphering the shared mechanisms of Gegen Qinlian Decoction in treating type 2 diabetes and ulcerative colitis via bioinformatics and machine learning.

Although previous clinical studies and animal experiments have demonstrated the efficacy of Gegen Qinlian Decoction (GQD) in treating Type 2 Diabetes Mellitus (T2DM) and Ulcerative Colitis (UC), the underlying mechanisms of its therapeutic effects remain elusive. This study aims to investigate the shared pathogenic mechanisms between T2DM and UC and elucidate the mechanisms through which GQD modulates these diseases using bioinformatics approaches. Data for this study were sourced from the Gene Expression Omnibus (GEO) database. Targets of GQD were identified using PharmMapper and SwissTargetPrediction, while targets associated with T2DM and UC were compiled from the DrugBank, GeneCards, Therapeutic Target Database (TTD), DisGeNET databases, and differentially expressed genes (DEGs). Our analysis encompassed six approaches: weighted gene co-expression network analysis (WGCNA), immune infiltration analysis, single-cell sequencing analysis, machine learning, DEG analysis, and network pharmacology. Through GO and KEGG analysis of weighted gene co-expression network analysis (WGCNA) modular genes and DEGs intersection, we found that the co-morbidity between T2DM and UC is primarily associated with immune-inflammatory pathways, including IL-17, TNF, chemokine, and toll-like receptor signaling pathways. Immune infiltration analysis supported these findings. Three distinct machine learning studies identified IGFBP3 as a biomarker for GQD in treating T2DM, while BACE2, EPHB4, and EPHA2 emerged as biomarkers for GQD in UC treatment. Network pharmacology revealed that GQD treatment for T2DM and UC mainly targets immune-inflammatory pathways like Toll-like receptor, IL-17, TNF, MAPK, and PI3K-Akt signaling pathways. This study provides insights into the shared pathogenesis of T2DM and UC and clarifies the regulatory mechanisms of GQD on these conditions. It also proposes novel targets and therapeutic strategies for individuals suffering from T2DM and UC.

Quantifying variation across 16S rRNA gene sequencing runs in human microbiome studies

Recent microbiome research has incorporated a higher number of samples through more participants in a study, longitudinal studies, and metanalysis between studies. Physical limitations in a sequencing machine can result in samples spread across sequencing runs. Here we present the results of sequencing nearly 1000 16S rRNA gene sequences in fecal (stabilized and swab) and oral (swab) samples from multiple human microbiome studies and positive controls that were conducted with identical standard operating procedures. Sequencing was performed in the same center across 18 different runs. The simplified mock community showed limitations in accuracy, while precision (e.g., technical variation) was robust for the mock community and actual human positive control samples. Technical variation was the lowest for stabilized fecal samples, followed by fecal swab samples, and then oral swab samples. The order of technical variation stability was inverse of DNA concentrations (e.g., highest in stabilized fecal samples), highlighting the importance of DNA concentration in reproducibility and urging caution when analyzing low biomass samples. Coefficients of variation at the genus level also followed the same trend for lower variation with higher DNA concentrations. Technical variation across both sample types and the two human sampling locations was significantly less than the observed biological variation. Overall, this research providing comparisons between technical and biological variation, highlights the importance of using positive controls, and provides semi-quantified data to better understand variation introduced by sequencing runs.Key points• Mock community and positive control accuracy were lower than precision.• Samples with lower DNA concentration had increased technical variation across sequencing runs.• Biological variation was significantly higher than technical variation due to sequencing runs.

Large-scale dynamic surgical scheduling under uncertainty by hierarchical reinforcement learning

Dynamic surgical scheduling within a workday is a complicated decision-making process. The critical challenge is that the actual duration of surgery and the arrival process of emergency patients are uncertain and unknown in advance. In this work, we propose a two-level dynamic scheduling framework based on hierarchical reinforcement learning to solve dynamic surgical scheduling problems considering both elective and emergency patients. Specifically, with the realisation of uncertainty, the upper-level agent (UA) dynamically decides whether to trigger rescheduling to optimise the workday total cost. The lower-level agent (LA) aims at obtaining subscheduling solutions when rescheduling is triggered. The subproblem at the LA can be formulated as a mixed integer programming model, which can be generalised to unrelated parallel machine scheduling with machine eligibility restrictions and sequence- and machine-dependent setup times. Such problems can be solved in small-scale cases and suffers the combinatorial explosion in large scale cases. To address this issue, we propose a heuristic method that is built upon deep reinforcement learning to obtain high-quality solutions. We conduct extensive simulation experiments with real data to test the effective of our framework. The results for different scenarios show that our proposed framework outperforms existing methods in terms of overall performance and has strong generalisation ability.

Unravelling the metabolic landscape of cutaneous melanoma: Insights from single-cell sequencing analysis and machine learning for prognostic assessment of lactate metabolism.

This manuscript presents a comprehensive investigation into the role of lactate metabolism-related genes as potential prognostic markers in skin cutaneous melanoma (SKCM). Bulk-transcriptome data from The Cancer Genome Atlas (TCGA) and GSE19234, GSE22153, and GSE65904 cohorts from GEO database were processed and harmonized to mitigate batch effects. Lactate metabolism scores were assigned to individual cells using the 'AUCell' package. Weighted Co-expression Network Analysis (WGCNA) was employed to identify gene modules correlated with lactate metabolism. Machine learning algorithms were applied to construct a prognostic model, and its performance was evaluated in multiple cohorts. Immune correlation, mutation analysis, and enrichment analysis were conducted to further characterize the prognostic model's biological implications. Finally, the function of key gene NDUFS7 was verified by cell experiments. Machine learning resulted in an optimal prognostic model, demonstrating significant prognostic value across various cohorts. In the different cohorts, the high-risk group showed a poor prognosis. Immune analysis indicated differences in immune cell infiltration and checkpoint gene expression between risk groups. Mutation analysis identified genes with high mutation loads in SKCM. Enrichment analysis unveiled enriched pathways and biological processes in high-risk SKCM patients. NDUFS7 was found to be a hub gene in the protein-protein interaction network. After the expression of NDUFS7 was reduced by siRNA knockdown, CCK-8, colony formation, transwell and wound healing tests showed that the activity, proliferation and migration of A375 and WM115 cell lines were significantly decreased. This study offers insights into the prognostic significance of lactate metabolism-related genes in SKCM.