mCNN-ETC: identifying electron transporters and their functional families by using multiple windows scanning techniques in convolutional neural networks with evolutionary information of protein sequences.

In the past decade, convolutional neural networks (CNNs) have been used as powerful tools by scientists to solve visual data tasks. However, many efforts of convolutional neural networks in solving protein function prediction and extracting useful information from protein sequences have certain limitations. In this research, we propose a new method to improve the weaknesses of the previous method. mCNN-ETC is a deep learning model which can transform the protein evolutionary information into image-like data composed of 20 channels, which correspond to the 20 amino acids in the protein sequence. We constructed CNN layers with different scanning windows in parallel to enhance the useful pattern detection ability of the proposed model. Then we filtered specific patterns through the 1-max pooling layer before inputting them into the prediction layer. This research attempts to solve a basic problem in biology in terms of application: predicting electron transporters and classifying their corresponding complexes. The performance result reached an accuracy of 97.41%, which was nearly 6% higher than its predecessor. We have also published a web server on http://bio219.bioinfo.yzu.edu.tw, which can be used for research purposes free of charge.

Consistency in human papillomavirus type detection between self-collected vaginal specimens and physician-sampled cervical specimens.

With the rising need for accessible cervical cancer screening, self-sampling methods offer a promising alternative to traditional physician-led sampling. This study aims to evaluate the efficacy of the HygeiaTouch Self Sampling Kit for Women in detecting human papillomavirus (HPV) types and predicting cervical lesions. We studied the concordance in identifying high-risk HPV (hrHPV) types between samples collected by physicians and those self-collected by women using a self-sampling kit for validation. Women aged 21-65, fitting into specific categories based on their cervical health history were eligible. Cohen's kappa coefficient to gauge concordance between the two specimen types and relative accuracy metrics in identifying cervical intraepithelial neoplasia (CIN) were also calculated, with physician-sampled specimens serving as a reference. A total of 1210 participants from three institutes were involved. The self-sampling kit closely matched the physician-led method in terms of collecting valid specimens (100% vs. 100%), identifying hrHPV types (kappa: 0.75, 95% confidence interval [95% CI]: 0.72-0.79; agreement: 87.7%, 95% CI: 85.8-89.6) and predicting CIN grade 2 or worse (CIN2+) (relative sensitivity: 0.949, relative accuracy: 0.959). Kappa values varied between 0.71 and 0.83 for different hrHPV types and combinations, with an overall value 0.75 (95% CI: 0.72-0.79) signifying robust compatibility between the two methods. Our study underscores the potential of the HygeiaTouch Self Sampling Kit as a reliable, efficient, and user-friendly alternative to traditional sampling methods. This suggests that self-sampling could be pivotal in expanding cervical cancer screening accessibility and enhancing detection rates.

Integrating Pre-Trained protein language model and multiple window scanning deep learning networks for accurate identification of secondary active transporters in membrane proteins.

Secondary active transporters play pivotal roles in regulating ion and molecule transport across cell membranes, with implications in diseases like cancer. However, studying transporters via biochemical experiments poses challenges. We propose an effective computational approach to identify secondary active transporters from membrane protein sequences using pre-trained language models and deep learning neural networks. Our dataset comprised 290 secondary active transporters and 5,420 other membrane proteins from UniProt. Three types of features were extracted - one-hot encodings, position-specific scoring matrix profiles, and contextual embeddings from the ProtTrans language model. A multi-window convolutional neural network architecture scanned the ProtTrans embeddings using varying window sizes to capture multi-scale sequence patterns. The proposed model combining ProtTrans embeddings and multi-window convolutional neural networks achieved 86% sensitivity, 99% specificity and 98% overall accuracy in identifying secondary active transporters, outperforming conventional machine learning approaches. This work demonstrates the promise of integrating pre-trained language models like ProtTrans with multi-scale deep neural networks to effectively interpret transporter sequences for functional analysis. Our approach enables more accurate computational identification of secondary active transporters, advancing membrane protein research.

Casuattimines A-N, fourteen new Lycopodium alkaloids from Lycopodiastrum casuarinoides with Cav3.1 channel inhibitory activity.

Two new dimeric Lycopodium alkaloids, casuattimines A and B (1 and 2), along with twelve previously undescribed Lycopodium alkaloids, casuattimines C-N (3-14), and eight known Lycopodium alkaloids, were isolated from Lycopodiastrum casuarinoides. Casuattimines A and B (1 and 2) are the first two ether-linked Lycopodium alkaloid dimers. Casuattimines C and D (3 and 4) are unique Lycopodium alkaloids characterized by a long fatty acid chain. Structural elucidation was achieved through HRESIMS, NMR, and electronic circular dichroism (ECD) calculations. In addition, the absolute configurations of compounds 7, 13, and 14 were determined by single crystal X-ray diffraction. Compounds 1, 2, and 4 demonstrated notable Cav3.1 channel inhibitory activities presenting IC50 values of 10.75 ± 1.02 µM, 9.33 ± 0.79 µM, and 7.14 ± 0.86 µM, respectively. The dynamics of compound 4 against the Cav3.1 channel and preliminary structure-activity relationships of these active Lycopodium alkaloids were also discussed.

Nephrotoxicity of organophosphate flame retardants in patients with chronic kidney disease: A 2-year longitudinal study.

Humans are extensively exposed to organophosphate flame retardants (OPFRs), an emerging group of organic contaminants with potential nephrotoxicity. Nevertheless, the estimated daily intake (EDI) and prognostic impacts of OPFRs have not been assessed in individuals with chronic kidney disease (CKD). In this 2-year longitudinal study of 169 patients with CKD, we calculated the EDIs of five OPFR triesters from urinary biomonitoring data of their degradation products and analyzed the effects of OPFR exposure on adverse renal outcomes and renal function deterioration. Our analysis demonstrated universal OPFR exposure in the CKD population, with a median EDIΣOPFR of 360.45 ng/kg body weight/day (interquartile range, 198.35-775.94). Additionally, our study revealed that high tris(2-chloroethyl) phosphate (TCEP) exposure independently correlated with composite adverse events and composite renal events (hazard ratio [95 % confidence interval; CI]: 4.616 [1.060-20.096], p = 0.042; 3.053 [1.075-8.674], p = 0.036) and served as an independent predictor for renal function deterioration throughout the study period, with a decline in estimated glomerular filtration rate of 4.127 mL/min/1.73 m2 (95 % CI, -8.127--0.126; p = 0.043) per log ng/kg body weight/day of EDITCEP. Furthermore, the EDITCEP and EDIΣOPFR were positively associated with elevations in urinary 8-hydroxy-2'-deoxyguanosine and kidney injury molecule-1 during the study period, indicating the roles of oxidative damage and renal tubular injury in the nephrotoxicity of OPFR exposure. To conclude, our findings highlight the widespread OPFR exposure and its possible nephrotoxicity in the CKD population.

Evaluating resectoscopy and the levonorgestrel intrauterine system for intermenstrual bleeding associated with cesarean scar defects: A retrospective study of treatment outcomes.

AIM: To compare and evaluate the efficacy of the levonorgestrel-releasing intrauterine system (LNG-IUD) and resectoscopy remodeling procedure for intermenstrual bleeding associated with previous cesarean delivery scar defect (PCDS). METHODS: A retrospective comparative study was conducted on patients with PCDS receiving LNG-IUD (levonorgestrel 20 µg/24 h, N = 33) or resectoscopy remodeling (N = 27). Treatment outcomes were compared over 1, 6, and 12 months. Outcomes in patients with a retroverted or large uterus size, defect size, and local vascularization also were evaluated. RESULTS: At 12 months post-treatment, there were no significant differences between groups in efficacy rate; however, the reduction of intermenstrual bleeding days was higher in the LNG-IUD group than in the resectoscopy group (13.6 vs. 8.5 days, p = 0.015). Within the first year, both groups experienced a reduction in bleeding days, but the decrease was greater in the LNG-IUD group. Individuals exhibiting increased local vascularization at the defect site experienced more favorable outcomes in the LNG-IUD group than the resectoscopy group (p = 0.016), and who responded poorly tended to have a significantly larger uterus in the LNG-IUD group (p = 0.019). No significant differences were observed in treatment outcomes for patients with a retroverted uterus or large defect in either group. CONCLUSIONS: Our findings support that the LNG-IUD is as effective as resectoscopy in reducing intermenstrual bleeding days associated with PCDS and can be safely applied to patients without recent fertility aspirations. Patients with increased local vascularization observed during hysteroscopy may benefit more from LNG-IUD intervention than resectoscopy.

Lycopodium Alkaloids from Huperzia goebelii.

One new fawcettimine-type Lycopodium alkaloid, hupertimine F (1), together with five known (2-6) Lycopodium alkaloids were isolated from Huperzia goebelii. The structure of 1 was elucidated by 1D and 2D NMR spectra, HRESIMS, and X-ray diffraction. Structurally, 1 represents the fourth example of Lycopodium alkaloids characterized by a 5/5/5/5/6 pentacyclic ring system with a 1-aza-7-oxabicyclo[2.2.1]heptane moiety. These known compounds 2, 3, 5, and 6 were isolated from H. goebelii for the first time. Compounds 1-6 were evaluated for acetylcholinesterase, butyrylcholinesterase and monoamine oxidase B inhibitory activities in vitro.

Molecular Characteristics and Virulence Profile of Clinical Listeria monocytogenes Isolates in Northern Taiwan, 2009-2019.

Listeria monocytogenes is a critical foodborne pathogen that causes severe invasive and noninvasive diseases and is associated with high mortality. Information on the prevalence of L. monocytogenes infections in Taiwan is very limited. This study aimed to analyze the molecular epidemiological surveillance and virulence gene distribution of 176 human clinical L. monocytogenes isolates collected between 2009 and 2019 in northern Taiwan. Our results showed that the isolates belonged to 4 serogroups (IIa, IIb, IVb, and IIc), with most isolates in serogroups IIa (81/176, 46%) and IIb (71/176, 40.3%). Multilocus sequence typing analysis revealed 18 sequence types (STs) and 13 clonal complexes (CCs). Eighty-four percent of all isolates belonged to six STs: CC87-ST87 (40/176, 22.7%), CC19-ST378 (36/176, 19.9%), CC155-ST155 (28/176, 15.5%), CC1-ST710 (16/176, 8.8%), CC5-ST5 (16/176, 8.8%), and CC101-ST101 (11/176, 6.1%). Furthermore, our analysis showed the distributions of four Listeria pathogenicity islands (LIPI) among all isolates. LIPI-1 and LIPI-2 existed in all isolates, whereas LIPI-3 and LIPI-4 only existed in specific STs and CCs. LIPI-3 existed in the STs, CC1-ST710, CC3-ST3, CC288-ST295, and CC191-ST1458, whereas LIPI-4 could be found in the STs, CC87-ST87 and CC87-ST1459. Strains containing LIPI-3 and LIPI-4 are potentially hypervirulent; thus, 68/176 isolates (39.1%) collected in this study were potentially hypervirulent. Since L. monocytogenes infections are considered highly correlated with diet, molecular epidemiological surveillance of Listeria in food is important; continued surveillance will provide critical information to prevent foodborne diseases.

Real-world efficacy and safety of low-dose pembrolizumab in patients with advanced and refractory gynecologic cancers.

OBJECTIVE: The approved standard dose of pembrolizumab (200 mg administrated every 3 weeks) for cancer treatment imposes a significant financial burden on patients. However, no study has analyzed the clinical outcomes of low-dose pembrolizumab among individuals diagnosed with gynecologic cancer. The primary objective of this study was to assess the effectiveness and safety of a low-dose pembrolizumab regimen in real-world clinical practice. METHODS: We retrospectively assessed the efficacy and safety data of patients with gynecologic malignancies who received pembrolizumab between 2017 and 2022 at Kaohsiung Chang Gung Memorial Hospital. Furthermore, we conducted a comparative analysis of the objective response rate (ORR) and progression-free survival (PFS) between patients with deficient mismatch repair (dMMR) and proficient MMR (pMMR). RESULTS: A total of thirty-nine patients were included and received pembrolizumab at fixed dosages of 50 mg (5.1%), 100 mg (84.6%) and 200 mg (10.3%) per cycle. Compared to the pMMR group, the dMMR group exhibited a tendency toward improved ORR (45.5% vs. 13.0%, p = 0.074), and notably, the median duration of response remained unreached. There was no significant difference in PFS between the dMMR and pMMR groups; however, the patients with dMMR in tumor tissue had a trend of better survival (p = 0.079). Incidence of immune-related adverse events (irAEs) of any grade was observed in 13 patients (33.3%), with 3 individuals (7.7%) experiencing grade 3 or 4 events. CONCLUSION: Low-dose pembrolizumab may be a cost-effective and safe treatment option without compromising clinical outcomes in patients with refractory gynecologic cancers.

Rational Lithium Salt Molecule Tuning for Fast Charging/Discharging Lithium Metal Battery.

The electrolytes for lithium metal batteries (LMBs) are plagued by a low Li+ transference number (T+) of conventional lithium salts and inability to form a stable solid electrolyte interphase (SEI). Here, we synthesized a self-folded lithium salt, lithium 2-[2-(2-methoxy ethoxy)ethoxy]ethanesulfonyl(trifluoromethanesulfonyl) imide (LiETFSI), and comparatively studied with its structure analogue, lithium 1,1,1-trifluoro-N-[2-[2-(2-methoxyethoxy)ethoxy)]ethyl]methanesulfonamide (LiFEA). The special anion chemistry imparts the following new characteristics: i) In both LiFEA and LiETFSI, the ethylene oxide moiety efficiently captures Li+, resulting in a self-folded structure and high T+ around 0.8. ii) For LiFEA, a Li-N bond (2.069 Å) is revealed by single crystal X-ray diffraction, indicating that the FEA anion possesses a high donor number (DN) and thus an intensive interphase "self-cleaning" function for an ultra-thin and compact SEI. iii) Starting from LiFEA, an electron-withdrawing sulfone group is introduced near the N atom. The distance of Li-N is tuned from 2.069 Šin LiFEA to 4.367 Šin LiETFSI. This alteration enhances ionic separation, achieves a more balanced DN, and tunes the self-cleaning intensity for a reinforced SEI. Consequently, the fast charging/discharging capability of LMBs is progressively improved. This rationally tuned anion chemistry reshapes the interactions among Li+, anions, and solvents, presenting new prospects for advanced LMBs.

Recent advances in features generation for membrane protein sequences: From multiple sequence alignment to pre-trained language models.

Membrane proteins play a crucial role in various cellular processes and are essential components of cell membranes. Computational methods have emerged as a powerful tool for studying membrane proteins due to their complex structures and properties that make them difficult to analyze experimentally. Traditional features for protein sequence analysis based on amino acid types, composition, and pair composition have limitations in capturing higher-order sequence patterns. Recently, multiple sequence alignment (MSA) and pre-trained language models (PLMs) have been used to generate features from protein sequences. However, the significant computational resources required for MSA-based features generation can be a major bottleneck for many applications. Several methods and tools have been developed to accelerate the generation of MSAs and reduce their computational cost, including heuristics and approximate algorithms. Additionally, the use of PLMs such as BERT has shown great potential in generating informative embeddings for protein sequence analysis. In this review, we provide an overview of traditional and more recent methods for generating features from protein sequences, with a particular focus on MSAs and PLMs. We highlight the advantages and limitations of these approaches and discuss the methods and tools developed to address the computational challenges associated with features generation. Overall, the advancements in computational methods and tools provide a promising avenue for gaining deeper insights into the function and properties of membrane proteins, which can have significant implications in drug discovery and personalized medicine.

Physicochemical Properties and Route of Systemic Delivery Control the In Vivo Dynamics and Breakdown of Radiolabeled Gold Nanostars.

The in vivo dynamics of nanoparticles requires a mechanistic understanding of multiple factors. Here, for the first time, the surprising breakdown of functionalized gold nanostars (F-AuNSs) conjugated with antibodies and 64 Cu radiolabels in vivo and in artificial lysosomal fluid ex vivo, is shown. The short-term biodistribution of F-AuNSs is driven by the route of systemic delivery (intravenous vs intraperitoneal) and long-term fate is controlled by the tissue type in vivo. In vitro studies including endocytosis pathways, intracellular trafficking, and opsonization, are combined with in vivo studies integrating a milieu of spectroscopy and microcopy techniques that show F-AuNSs dynamics is driven by their physicochemical properties and route of delivery. F-AuNSs break down into sub-20 nm broken nanoparticles as early as 7 days postinjection. Martini coarse-grained simulations are performed to support the in vivo findings. Simulations suggest that shape, size, and charge of the broken nanoparticles, and composition of the lipid membrane depicting various tissues govern the interaction of the nanoparticles with the membrane, and the rate of translocation across the membrane to ultimately enable tissue clearance. The fundamental study addresses critical gaps in the knowledge regarding the fate of nanoparticles in vivo that remain a bottleneck in their clinical translation.

Addressing data imbalance problems in ligand-binding site prediction using a variational autoencoder and a convolutional neural network.

Since 2015, a fast growing number of deep learning-based methods have been proposed for protein-ligand binding site prediction and many have achieved promising performance. These methods, however, neglect the imbalanced nature of binding site prediction problems. Traditional data-based approaches for handling data imbalance employ linear interpolation of minority class samples. Such approaches may not be fully exploited by deep neural networks on downstream tasks. We present a novel technique for balancing input classes by developing a deep neural network-based variational autoencoder (VAE) that aims to learn important attributes of the minority classes concerning nonlinear combinations. After learning, the trained VAE was used to generate new minority class samples that were later added to the original data to create a balanced dataset. Finally, a convolutional neural network was used for classification, for which we assumed that the nonlinearity could be fully integrated. As a case study, we applied our method to the identification of FAD- and FMN-binding sites of electron transport proteins. Compared with the best classifiers that use traditional machine learning algorithms, our models obtained a great improvement on sensitivity while maintaining similar or higher levels of accuracy and specificity. We also demonstrate that our method is better than other data imbalance handling techniques, such as SMOTE, ADASYN, and class weight adjustment. Additionally, our models also outperform existing predictors in predicting the same binding types. Our method is general and can be applied to other data types for prediction problems with moderate-to-heavy data imbalances.

A transformer architecture based on BERT and 2D convolutional neural network to identify DNA enhancers from sequence information.

Recently, language representation models have drawn a lot of attention in the natural language processing field due to their remarkable results. Among them, bidirectional encoder representations from transformers (BERT) has proven to be a simple, yet powerful language model that achieved novel state-of-the-art performance. BERT adopted the concept of contextualized word embedding to capture the semantics and context of the words in which they appeared. In this study, we present a novel technique by incorporating BERT-based multilingual model in bioinformatics to represent the information of DNA sequences. We treated DNA sequences as natural sentences and then used BERT models to transform them into fixed-length numerical matrices. As a case study, we applied our method to DNA enhancer prediction, which is a well-known and challenging problem in this field. We then observed that our BERT-based features improved more than 5-10% in terms of sensitivity, specificity, accuracy and Matthews correlation coefficient compared to the current state-of-the-art features in bioinformatics. Moreover, advanced experiments show that deep learning (as represented by 2D convolutional neural networks; CNN) holds potential in learning BERT features better than other traditional machine learning techniques. In conclusion, we suggest that BERT and 2D CNNs could open a new avenue in biological modeling using sequence information.

Estrogen/Progesterone Receptor Expression and Cancer Antigen 125 Level as Preoperative Predictors to Estimate Lymph Node Metastasis in Endometrioid Endometrial Cancer.

Loss of estrogen receptor/progesterone receptor (ER/PR) in endometrial cancer (EC) is associated with tumor progression and poor outcomes. Elevated pretreatment cancer antigen 125 (CA 125) level is a risk factor for lymph node metastasis (LNM). We evaluated whether the combination of ER/PR expression and CA 125 level could be used as a biomarker to predict LNM. We retrospectively investigated patients with endometrioid EC who underwent complete staging surgery during January 2015 to December 2020. We analyzed ER/PR status using immunohistochemical staining, and quantified its expression using the sum of both ER/PR H-scores. Receiver operating characteristic curves were used to identify optimal cutoff values of H-score and CA 125 levels for predicting LNM. A nomogram for predicting LNM was constructed and validated by bootstrap resampling. In 396 patients, the optimal cutoff values of the ER/PR H-score and CA 125 were 407 (area under the receiver operating characteristic curve: 0.645, P=0.001) and 40 U/mL (area under the receiver operating characteristic curve: 0.762, P<0.001), respectively. Multivariate analysis showed that CA 125 ≥40 UmL (odds ratio: 10.02; 95% CI: 4.74-21.18) and ER/PR H-score <407 (odds ratio: 4.20; 95% CI: 1.55-11.32) were independent predictors. An LNM predictive nomogram was constructed using these 2 variables and our model yielded a negative predictive value and negative likelihood ratio of 98.3% and 0.14, respectively. ER/PR expression with pretreatment CA 125 levels can help estimate LNM risk and aid in decision-making regarding the need for lymphadenectomy in patients with endometrioid EC.

Correlation between the dopaminergic system and inflammation disease: a review.

The dopaminergic system is inextricably linked with neurological diseases and addiction. In recent years, many studies have found that the dopaminergic system involves in inflammatory diseases, particularly neuroinflammatory diseases development; This review summarizes the studies of dopaminergic system in inflammatory diseases, and specifically highlights the mechanisms of how dopaminergic system regulates inflammation; In addition, we speculate that there are some cavities in current research, including mixed usage of inhibitors, agonists and lack of systematic controls; We expect this review would provide directions to future research of dopaminergic system and inflammatory diseases.

Clonal Spreading of ST42 Staphylococcus haemolyticus Strains Occurs Possibly Due to fusB and tetK Resistant Genes and Capsule-Related Genes.

Multi-drug resistant Staphylococcus haemolyticus is a frequent nosocomial invasive bacteremia pathogen in hospitals. Our previous analysis showed one of the predominant strains, ST42 originated from ST3, had only one multilocus sequence typing (MLST) variation among seven loci in SH1431; yet no significant differences in biofilm formation observed between ST42 and ST3, suggesting that other factors influence clonal lineage change. Whole genome sequencing was conducted on two isolates from ST42 and ST3 to find phenotypic and genotypic variations, and these variations were further validated in 140 clinical isolates. The fusidic acid- and tetracycline-resistant genes (fusB and tetK) were found only in CGMH-SH51 (ST42). Further investigation revealed consistent resistant genotypes in all isolates, with 46% and 70% of ST42 containing fusB and tetK, respectively. In contrast, only 23% and 4.2% ST3 contained these two genes, respectively. The phenotypic analysis also showed that ST42 isolates were highly resistant to fusidic acid (47%) and tetracycline (70%), compared with ST3 (23% and 4%, respectively). Along with drug-resistant genes, three capsule-related genes were found in higher percentage distributions in ST42 than in ST3 isolates. Our findings indicate that ST42 could become endemic in Taiwan, further constitutive surveillance is required to prevent the spread of this bacterium.

Supramolecular Polymer Ion Conductor with Weakened Li Ion Solvation Enables Room Temperature All-Solid-State Lithium Metal Batteries.

Improved durability, enhanced interfacial stability, and room temperature applicability are desirable properties for all-solid-state lithium metal batteries (ASSLMBs), yet these desired properties are rarely achieved simultaneously. Here, in this work, it is noticed that the huge resistance at Li metal/electrolyte interface dominantly impeded the normal cycling of ASSLMBs especially at around room temperature (<30 °C). Accordingly, a supramolecular polymer ion conductor (SPC) with "weak solvation" of Li+ was prepared. Benefiting from the halogen-bonding interaction between the electron-deficient iodine atom (on 1,4-diiodotetrafluorobenzene) and electron-rich oxygen atoms (on ethylene oxide), the O-Li+ coordination was significantly weakened. Therefore, the SPC achieves rapid Li+ transport with high Li+ transference number, and importantly, derives a unique Li2 O-rich SEI with low interfacial resistance on lithium metal surface, therefore enabling stable cycling of ASSLMBs even down to 10 °C. This work is a new exploration of halogen-bonding chemistry in solid polymer electrolyte and highlights the importance of "weak solvation" of Li+ in the solid-state electrolyte for room temperature ASSLMBs.

Using multiple convolutional window scanning of convolutional neural network for an efficient prediction of ATP-binding sites in transport proteins.

Protein multiple sequence alignment information has long been important features to know about functions of proteins inferred from related sequences with known functions. It is therefore one of the underlying ideas of Alpha fold 2, a breakthrough study and model for the prediction of three-dimensional structures of proteins from their primary sequence. Our study used protein multiple sequence alignment information in the form of position-specific scoring matrices as input. We also refined the use of a convolutional neural network, a well-known deep-learning architecture with impressive achievement on image and image-like data. Specifically, we revisited the study of prediction of adenosine triphosphate (ATP)-binding sites with more efficient convolutional neural networks. We applied multiple convolutional window scanning filters of a convolutional neural network on position-specific scoring matrices for as much as useful information as possible. Furthermore, only the most specific motifs are retained at each feature map output through the one-max pooling layer before going to the next layer. We assumed that this way could help us retain the most conserved motifs which are discriminative information for prediction. Our experiment results show that a convolutional neural network with not too many convolutional layers can be enough to extract the conserved information of proteins, which leads to higher performance. Our best prediction models were obtained after examining them with different hyper-parameters. Our experiment results showed that our models were superior to traditional use of convolutional neural networks on the same datasets as well as other machine-learning classification algorithms.

Comprehensive Clinicopathologic Analysis for Mismatch Repair Protein Expression in Unselected Endometrial Carcinoma Patients With an Emphasis on the Role of MLH1 Deficiency.

Screening for mismatch repair (MMR) deficiency in unselected patients with endometrial carcinoma (EC) and the clinicopathologic descriptions of ECs with MMR deficiency have been well demonstrated in Western populations, but studies on Asian populations are relatively scarce. In this study, we described the clinicopathologic features of ECs according to MMR status in unselected Taiwanese patients. We also conducted subgroup analysis of MMR-deficient (dMMR) cases according to the presence or absence of MLH1. Patients diagnosed with ECs between January 2017 and February 2020 at our institution were included. Immunohistochemistry analysis of MLH1, PMS2, MSH2, and MSH6 proteins on endometrial primary tumors and clinicopathologic variables were assessed retrospectively. A total of 231 EC patients were enrolled, of whom 50 (21.6%) had dMMR tumors. Of these 50 cases, 39 had tumors that lacked MLH1 expression and 11 were positive for MLH1. The overall dMMR group was significantly related to older age, parity, and high histologic grade compared with the MMR-proficient (pMMR) group. ECs with MLH1 deficiency were obviously associated with several poor pathologic features, including high histologic grade, lymph node metastasis, and lymphovascular space invasion. Moreover, we first reported that parity and the late age at menopause are strongly correlated with MLH1-related dMMR EC group compared with pMMR group. In conclusion, triaging EC patients into pMMR, MLH1-related dMMR and non-MLH1-related dMMR groups by immunohistochemistry analysis may help clinicians to predict disease behavior and guide further management. The strong association between parity and MLH1-related dMMR ECs warrants further investigation on the underlying mechanism.