[Clinicopathological characteristics and molecular alterations of primary cardiac leiomyosarcoma: report of five cases].

Objective: To investigate the clinical, pathologic and radiologic features and molecular alterations in patients with primary cardiac leiomyosarcoma (PCLMS). Methods: Five cases of PCLMS were collected in Beijing Anzhen Hospital from January 2016 to December 2020. The clinical, pathologic and radiologic data, and molecular alterations were analyzed, and the patients were followed up. Results: All five patients were female, and had no history of leiomyosarcoma in other parts of the body. The age of patients ranged from 37 to 62 years (median 47 years). The main clinical symptoms were chest pain and dyspnea, one also presented with palpitation and lower limb weakness and one with dizziness. Two tumors were located in the left atrium, two in the right atrium, and one in the right ventricle, and they maximal diameter ranged from 2.5 to 14.0 cm (mean 6.2 cm). The neoplasms presented as medium-echo masses with a broad base in the echocardiography, and as a low-density, solid mass when detected by contrast-enhanced CT. Histologically, two tumors were well-differentiated and three were moderately and poorly differentiated, and two included extensive, loose myxoid stroma. Immunohistochemical staining showed that PCLMS was positive for SMA, desmin, MDM2, and epidermal growth factor receptor. Fluorescence in situ hybridization showed ALK gene rearrangement in two cases, and COL1A1-PDGFB fusion in three cases. All cases received surgical excision and two cases received chemotherapy. Three patients died within 0-11 months (mean survival of 7.7 months) and two patients were alive. Conclusions: PCLMS is a malignant tumor with a high recurrence rate and poor prognosis. These cases may provide useful information to improve the diagnosis and management of PCLMS.

[A meta-analysis of risk factors for multidrug-resistant tuberculosis in China].

Objective: To explore the main risk factors of multidrug-resistant tuberculosis (MDR-TB) in China and to provide evidence-based evidence for MDR-TB preventon and control. Methods: All relevant literatures were searched in thedatabases, such as Pubmed, Web of Science and CNKI, Wanfang, VIP and SinoMed from 2000 to 2021. Quality evaluation and data extraction were carried out, and then a meta-analysis was performed using Stata 16.0 software. Results: A total of 59 literatures (36 cross-sectional and 23 case-control) including 75 793 participants were included in this study, and meta-analysis results showed age (OR=1.27, 95%CI: 1.05-1.54), education level (OR=1.29, 95%CI: 1.02-1.65), positive sputum smear (OR=2.56, 95%CI: 1.09-6.04), pulmonary cavity (OR=1.99, 95%CI: 1.57-2.52), course of disease (OR=4.25, 95%CI: 1.95-9.30), history of tuberculosis treatment (OR=6.42,95%CI:5.40-7.63), treatment interruption (OR=2.81, 95%CI: 1.50-5.29), irregular medication (OR=5.02, 95%CI: 2.95-8.54), adverse drug reactions (OR=4.27, 95%CI: 2.22-8.19), combined chronic obstructive pulmonary disease (COPD) (OR=2.21, 95%CI: 1.45-3.37), tuberculosis exposure history (OR=1.99, 95%CI: 1.36-2.91), smoking history (OR=1.35, 95%CI: 1.09-1.66) and floating population (OR=1.60, 95%CI: 1.04-2.44) were associated with the occurrence of MDR-TB. Conclusions: The high risk groups were farmer, low education level, pulmonary cavity, long course of disease, history of tuberculosis treatment, treatment interruption, irregular medication, adverse drug reaction, co-COPD, contact history of tuberculosis, smoking history, rural residence, and floating population. We should pay attention to high-risk groups, strengthen management and take effective measures such as early screening, knowledge education on tuberculosis, standardized and personalized treatment and whole-course supervision.

[Application of deep learning neural network in pathological image classification of non-inflammatory aortic membrane degeneration].

Objective: To investigate the value of deep learning in classifying non-inflammatory aortic membrane degeneration. Methods: Eighty-nine cases of non-inflammatory aortic media degeneration diagnosed from January to June 2018 were collected at Beijing Anzhen Hospital, Capital Medical University, China and scanned into digital sections. 1 627 hematoxylin and eosin stained photomicrographs were extracted. Combined with the ResNet18-based deep convolution neural network model, 4-category classification of pathological images were performed to diagnose the non-inflammatory aortic lesion. Results: The prediction model of artificial intelligence assisted diagnosis had the best accuracy, sensitivity and precision in identifying lesions with smooth muscle cell nuclei loss, which were 99.39%, 98.36% and 98.36%, respectively. The classification accuracy of elastic fiber fragmentation and/or loss lesions was 98.08%, while that of intralamellar mucoid extracellular matrix accumulation lesions was 96.93%. The overall accuracy of the classification model was 96.32%, and the area under the curve was 0.982. Conclusions: The accuracy of deep learning neural network model in the 4-category classification of non-inflammatory aortic lesionsis confirmed based on digital photomicrographs. This method can effectively improve the diagnostic efficiency of pathologists.

[The therapeutic effect of carnosine combined with dexamethasone in the lung injury of seawater-drowning].

Objective: To explore the therapeutic effect of carnosine and dexamethasone in lung injury caused by seawater drowning. Methods: The in vitro experiments with A549 cells were divided into 5 groups: blank control group (C), seawater injury group (S), seawater injury+dexamethasone treatment group (S+D), seawater injury+carnosine treatment group (S+C), seawater injury dexamethasone and carnosine combined therapy(S+D+C) group. The optimal therapeutic dose of drugs for the treatment of seawater drowning lung injury was tested in vitro. Based on the optimal dose, the levels of TNF-α and IL-6 in each group at different time points were detected at the cell level by ELISA. The level of apoptosis was detected by flow cytometry. The in vivo experiments with SD rats were randomly divided into 5 groups (n=8 each): blank control group (RC),seawater drowning injury group (RS),seawater drowning injury+dexamethasone treatment group (RSD),seawater drowning injury+carnosine treatment group (RSC),seawater drowning injury+dexamethasone+carnosine combined treatment group (RSDC). The animal model with seawater inhalation acute lung injury was made by intratracheal infusion (4 ml/kg). The pathological changes of the lungs were observed. The expression of superoxide dismutase (SOD) in each group was detected by Western blot. Results: The results of in vitro experiments showed significant increase of apoptosis after seawater injury. The normal cell rate in group C was 98.3% while the apoptosis rate was 1.7%. The normal cell in group S was 18.8%, and the apoptosis rate was 81% (P<0.01). TNF-α and IL-6 levels in group S increased to 180.25 ng/L and 61.56 ng/L, respectively, which were statistically significant compared with group C (P<0.01). After drug protection, apoptosis was reduced in S+D group, S+C group and S+D+C group, with apoptosis rates of 65.4%, 70.9% and 42.6%, respectively. The contents of TNF-α and IL-6 also decreased in the S+D+C group (P<0.01). The results of in vivo experiments showed obvious lung injury and disordered lung tissue structures in the RS group at 4 h after modeling. There was hemorrhage in the pulmonary interstitium and a large number of inflammatory cells. Results of western blot showed that the expression of SOD increased in the RS group. Compared with RS group, the treatment alleviated acute lung injury and decreased the expression level of SOD in RSD, RSC and RSDC groups (P<0.01). Conclusion: Dexamethasone and carnosine reduced the influence of seawater inhalation on the lung in the rat model. The positive effect of combination of these two drugs on lung injury caused by seawater inhalation was stronger than a single drug.

[Clinicopathologic study of cardiac myxofibrosarcomas].

Objective: To investigate the clinicopathologic features of cardiac myxofibrosarcomas. Methods: The clinical data, pathomorphologic and immunohistochemical features were evaluated in five cases of cardiac myxofibrosarcoma collected from January 2009 to December 2014, with relevant literature review. Results: Five patients with cardiac myxofibrosarcoma, including four women and one man [age range 39-61 years; mean (50.4±9.0) years] were included. All tumors were broadbased and located mainly in the left atrium, with one case extending through the atrial wall and pericardium to the left lower lung lobe. The morphological grade was low in one case, intermediate in one, and high in three. Using Fédération Nationale des Centres de Lutte Contre le Cancer (FNLCC) grading system, one case was grade 1 and four cases were grade 2. Immunohistochemical analysis revealed diffuse and strong expression for vimentin in all cases. Smooth muscle actin and muscle specific actin were variably expressed. Complete tumor excision was performed in one case, and tumor debulking was performed in the other four cases. Clinical follow-up was available in three cases. One patient with en bloc excision of the tumor mass survived 13 months and the other two with tumor debulking died one month after surgery. Conclusions: The most common location for cardiac myxofibrosarcoma is the left atrium. Some myxofibrosarcoma may be histologically bland and misdiagnosed as myxoma due to histological similarities. Thus caution should be exercised in their microscopic differentiation. Precise imaging, multidisciplinary approach and adequate initial surgery may contribute to improving the clinical outcomes of myxofibrosarcoma.

Association of the LMNA gene single nucleotide polymorphism rs4641 with bdilated cardiomyopathy.

Recently, studies on the pathogenesis of dilated cardiomyopathy (DCM) have focused on the underlying molecular biology and the association between single nucleotide polymorphisms (SNPs) and disease. This study was designed to explore the association between the rs4641 SNP of the LMNA gene and DCM in order to identify a new gene locus related to DCM. Polymerase chain reaction-restriction fragment length polymorphism and DNA sequencing were employed to detect and genotype rs4641 in 198 patients with DCM and 160 healthy controls. Genotype and allele frequencies were compared to discover their relationship and logistic regression was used to assess the risk of DCM associated with the polymorphic variants. In the DCM group, the frequencies of the TC and TT genotypes and the T allele of rs4641 were remarkably higher than those in the control group (P < 0.01). According to risk analysis, taking the CC genotype as a reference, both the TC and TT genotypes increased the risk of DCM pathogenesis, with OR (95%CI) values of 5.957 (2.903- 12.222) and 6.424 (2.156-19.141), respectively. Taking the C allele as the reference, presence of the T allele was found to increase DCM risk, with OR (95%CI) of 5.295 (3.121-8.983). These results suggested that the C to T mutation at the rs4641 locus of LMNA could enhance the risk of DCM, and that rs4641 represented a genetic susceptibility locus. Therefore, it was concluded that the LMNA rs4641 SNP was associated with DCM risk, which indicated that LMNA is a susceptibility gene for DCM.

Genetic analysis of health traits and their associations with longevity, fertility, production, and conformation traits in Holstein cattle.

Health traits have high economic values in dairy cattle breeding, which can cause considerable financial loss through involuntary culling. In this study, fourteen health traits were analysed, including five composite health traits: reproductive disorders, udder health (UH), digestive disorders, metabolic disorders, locomotory diseases (LD), and nine independent health traits: gestation disorders and peripartum disorders, irregular estrus cycle and sterility, metritis (ME), mastitis (MA), abomasal displacement (AD), enteritis (EN), and ketosis, claw diseases (CD), laminitis complex. This study analysed variance components for health traits through both single and bivariate repeatability animal models. All health traits showed low heritability, ranging from 0.001 to 0.025. Most of the health traits in five categories showed negative genetic correlations, ranging from -0.012 (CD and EN) to -0.634 (ME and EN). Strong positive genetic correlations appeared within the same category, ranging from 0.469 (EN and AD) to 0.994 (UH and MA, LD and CD). Furthermore, approximate genetic correlations were evaluated between health traits and routinely collected traits (longevity, fertility, production, and conformation). In general, the low to moderate approximate genetic correlations were estimated between health traits and routinely collected traits. The estimated correlations between health traits and longevity, fertility, production, and conformation traits could provide an indirect reference for disease-resistance breeding in Holstein cattle.

Genetic analysis of longevity and their associations with fertility traits in Holstein cattle.

The increase of longevity is intended to reduce involuntary culling rates, not extend the life span, and it reflects the ability of animals to successfully cope with the environment and disease during production. Sire model, animal model and repeatability animal models were used to estimate the (co) variance components of longevity and fertility traits. Six longevity and thirteen fertility traits were analysed, including herd life (HL), productive life (PL), number of days between first calving and the end of first lactation or culling (L1); number of days between first calving and the end of the second lactation or culling (L2); number of days between first calving and the end of the third lactation or culling (L3); number of days between first calving and the end of the fourth lactation or culling (L4); age at first service, age at first calving (AFC), the interval from first to last inseminations in heifer (IFLh), conception rate of first insemination in heifer, days open (DO), calving interval, gestation length, interval from calving to first insemination (ICF), interval from first to last inseminations in cow (IFLc), conception rate of first insemination in cow, calving ease (CE), birth weight, and calf survival. The estimated heritabilities (±SE) were 0.018 (±0.003), 0.015 (±0.003), 0.049 (±0.004), 0.025 (±0.003), 0.009 (±0.002) and 0.011 (±0.002) for HL, PL, L1, L2, L3 and L4, respectively. Strong correlations were appeared in HL and PL; the genetic and phenotypic correlation coefficients were 0.998 and 0.985, respectively. There were high genetic and phenotypic correlations which were observed in L1 and L2, L2 and L3, L3 and L4, respectively. All fertility traits of heifer showed medium to high heritability, while the cow showed low heritability. All heifer fertility traits had low genetic associations with longevity traits, ranging from -0.018 (L2 and IFLh) to 0.257 (L3 and AFC). Most of the fertility traits showed negative correlations with longevity traits in different parities, and we recommend DO, ICF, IFLc and CE as indirect indicators of longevity traits in dairy cows, but we also need to take into account the differences between parities.

[Predictions of achievement of Sustainable Development Goal to reduce age-standardized mortality rate of four major non-communicable diseases by 2030 in China].

Objective: To predicate whether China can achieve the United Nations Sustainable Development Goals (SDGs) 3.4.1 to reduce the age-standardized mortality rate of four major non-communicable diseases (NCDs) in residents aged 30-70 years by 2030 based on the trend of the mortality from 1990 to 2019. Methods: We collected the mortality data on cardiovascular diseases, cancers, chronic respiratory diseases, and diabetes by age, gender and year in China from the Global Disease Burden Study 2019 (GBD2019). The age-period-cohort (APC) Bayesian model was applied for modeling the age-standardized mortality rate of four major NCDs in China during 2020-2030 according to the trend of the mortality during 1990-2019, and comparing the predicted value in 2030 with the observed value in 2015 to evaluate the possibility of achieving SDGs 3.4.1. Results: The age-standardized mortality rate of the four major NCDs in China showed a downward trend during 1990-2019. It is predicted that the number of death of the four NCDs in Chinese residents aged 30-70 years would increase from 2.96 million in 2020 to 3.19 million in 2030, while the age-standardized mortality rate would decrease from 308.49/100 000 in 2020 to 277.80/100 000 in 2030. The age-standardized mortality rate in 2030 would only decrease by 15.94% (18.73% for males and 14.31% for females) compared with 330.46/100 000 in 2015, with a 25.09% decrease for cardiovascular diseases, 4.76% for cancers, 37.21% for chronic respiratory diseases, and unchanged for diabetes. Conclusion: Although the age-standardized mortality rate of four major NCDs declined from 1990 to 2019 in China, it is difficult to achieve the SDGs of a 1/3 mortality rate reduction by 2030 according to the current declining trend, suggesting more active and effective efforts for NCD prevention and control are needed.

Associations between single-nucleotide polymorphisms (+45T>G, +276G>T, -11377C>G, -11391G>A) of adiponectin gene and type 2 diabetes mellitus: a systematic review and meta-analysis.

AIMS/HYPOTHESIS: The associations between adiponectin polymorphisms and type 2 diabetes have been studied widely; however, results are inconsistent. METHODS: We searched electronic literature databases and reference lists of relevant articles. A fixed or random effects model was used on the basis of heterogeneity. Sub-group and meta-regression analyses were conducted to explore the sources of heterogeneity. RESULTS: There were no statistically significant associations between +45T>G (rs2241766), +276G>T (rs1501299), -11391G>A (rs17300539) and type 2 diabetes risk. However, for -11377C>G (rs266729), the pooled OR (95% CI) for G vs C allele was 1.07 (1.03-1.11, p = 0.001). Subgroup analysis by study design revealed that -11377C>G (rs266729) dominant model (CG+GG vs CC, p = 0.0008) and G vs C allele (p = 0.0004) might be associated with type 2 diabetes risk in population-based case-control studies. After stratification by ethnicity, we found that -11377C>G (rs266729) dominant model (CG+GG vs CC, p = 0.004) and G vs C allele (p = 0.001) might be associated with type 2 diabetes risk in white individuals. In individuals with a family history of diabetes, the presence of -11391G>A (rs17300539) dominant model (GA+AA vs GG) and A vs G allele might be associated with increased risk of type 2 diabetes. CONCLUSIONS/INTERPRETATION: The presence of +45T>G (rs2241766), +276G>T (rs1501299) and -11391G>A (rs17300539) do not appear to influence the development of type 2 diabetes. However, G vs C allele of -11377C>G (rs266729) might be a risk factor for type 2 diabetes.

Monodispersed NiO nanoflowers with anomalous magnetic behavior.

Nickel oxide (NiO) nanoflowers, prepared by thermal decomposition, exhibit anomalous magnetic properties far below the blocking temperature, i.e., a cusp in both the zero-field-cooled and field-cooled curves at about 21 K. Detailed characterization discloses that the individual NiO nanoflower consists of porous crystals with holes (1.0-1.5 nm in size) inside. We believe that the low temperature magnetic feature observed here could be a new kind of spin transition for the uncompensated spins around the holes and will trigger more studies in other nanostructured antiferromagnetic materials.

[Current status and projection of non-communicable diseases in 126 countries participating in the Belt and Road initiative].

Objective: To compare the indicators of non-communicable diseases (NCD) and predict the achieving time of United Nations (UN) Sustainable Development Goals (SDG) in 125 countries participating in the Belt and Road (B&R) initiative and China. Methods: Using the open access data of Global Burden of Disease study, we first got the premature mortality rates of four main chronic diseases (cardiovascular disease, cancer, diabetes and chronic respiratory diseases) and suicide mortality rate in the 126 countries from1990 to 2017. We transformed the value of each indicator into a scale of 0-100 in percentile for each country and applied geometric mean to calculate total NCD score for comparison among 126 countries. We then examined the association of NCD scores with socio-demographic index (SDI) values. Finally, we used annualized rates of change during 1990-2015 to predict achieving time of the UN goal by 2030 for each indicator of chronic diseases premature mortality rate and suicide mortality rates in each B&R country. Results: The integral median of total NCD score in the 126 countries in 2017 was 82.7. The score of China was 87.6, ranking 33(rd). The top three countries were Kuwait (98.1), Peru (97.5) and Italy (96.0). The last three countries were Papua New Guinea (28.9), Vanuatu (54.7) and Ukraine (58.0). The total NCD score showed positive correlation with SDI values (r=0.33) mainly due to chronic disease indicator (r=0.45). Fifteen countries will achieve the SDG goal of chronic disease premature mortality in or before 2030, but China will achieve it in 2038. Fifteen countries are expected to achieve the goal of suicide mortality, and China will acheive the goal ahead of schedule in 2024. Conclusions: The NCD rates varied widely among the countries along B&R. It is a challenge to achieve the SDG goal of chronic disease premature mortality rate by 2030 for China. In order to achieve the SDG goals by 2030, we should strengthen multilateral cooperation and complement each other's advantages, and reduce NCD mortality of people and improve people's health in countries along B&R.

PharmGED: Pharmacogenetic Effect Database.

Prediction and elucidation of pharmacogenetic effects is important for facilitating the development of personalized medicines. Knowledge of polymorphism-induced and other types of drug-response variations is needed for facilitating such studies. Although databases of pharmacogenetic knowledge, polymorphism and toxicogenomic information have appeared, some of the relevant data are provided in separate web-pages and in terms of relatively long descriptions quoted from literatures. To facilitate easy and quick assessment of the relevant information, it is helpful to develop databases that provide all of the information related to a pharmacogenetic effect in the same web-page and in brief descriptions. We developed a database, Pharmacogenetic Effect Database (PharmGED), for providing sequence, function, polymorphism, affected drugs and pharmacogenetic effects. PharmGED can be accessed at http://bidd.cz3.nus.edu.sg/phg/ free of charge for academic use. It currently contains 1825 entries covering 108 disease conditions, 266 distinct proteins, 693 polymorphisms, 414 drugs/ligands cited from 856 references.

Homology-free prediction of functional class of proteins and peptides by support vector machines.

Protein and peptide sequences contain clues for functional prediction. A challenge is to predict sequences that show low or no homology to proteins or peptides of known function. A machine learning method, support vector machines (SVM), has recently been explored for predicting functional class of proteins and peptides from sequence-derived properties irrespective of sequence similarity, which has shown impressive performance for predicting a wide range of protein and peptide classes including certain low- and non- homologous sequences. This method serves as a new and valuable addition to complement the extensively-used alignment-based, clustering-based, and structure-based functional prediction methods. This article evaluates the strategies, current progresses, reported prediction performances, available software tools, and underlying difficulties in using SVM for predicting the functional class of proteins and peptides.

A support vector machines approach for virtual screening of active compounds of single and multiple mechanisms from large libraries at an improved hit-rate and enrichment factor.

Support vector machines (SVM) and other machine-learning (ML) methods have been explored as ligand-based virtual screening (VS) tools for facilitating lead discovery. While exhibiting good hit selection performance, in screening large compound libraries, these methods tend to produce lower hit-rate than those of the best performing VS tools, partly because their training-sets contain limited spectrum of inactive compounds. We tested whether the performance of SVM can be improved by using training-sets of diverse inactive compounds. In retrospective database screening of active compounds of single mechanism (HIV protease inhibitors, DHFR inhibitors, dopamine antagonists) and multiple mechanisms (CNS active agents) from large libraries of 2.986 million compounds, the yields, hit-rates, and enrichment factors of our SVM models are 52.4-78.0%, 4.7-73.8%, and 214-10,543, respectively, compared to those of 62-95%, 0.65-35%, and 20-1200 by structure-based VS and 55-81%, 0.2-0.7%, and 110-795 by other ligand-based VS tools in screening libraries of >or=1 million compounds. The hit-rates are comparable and the enrichment factors are substantially better than the best results of other VS tools. 24.3-87.6% of the predicted hits are outside the known hit families. SVM appears to be potentially useful for facilitating lead discovery in VS of large compound libraries.

PROFEAT: a web server for computing structural and physicochemical features of proteins and peptides from amino acid sequence.

Sequence-derived structural and physicochemical features have frequently been used in the development of statistical learning models for predicting proteins and peptides of different structural, functional and interaction profiles. PROFEAT (Protein Features) is a web server for computing commonly-used structural and physicochemical features of proteins and peptides from amino acid sequence. It computes six feature groups composed of ten features that include 51 descriptors and 1447 descriptor values. The computed features include amino acid composition, dipeptide composition, normalized Moreau-Broto autocorrelation, Moran autocorrelation, Geary autocorrelation, sequence-order-coupling number, quasi-sequence-order descriptors and the composition, transition and distribution of various structural and physicochemical properties. In addition, it can also compute previous autocorrelations descriptors based on user-defined properties. Our computational algorithms were extensively tested and the computed protein features have been used in a number of published works for predicting proteins of functional classes, protein-protein interactions and MHC-binding peptides. PROFEAT is accessible at http://jing.cz3.nus.edu.sg/cgi-bin/prof/prof.cgi.

Prediction of MHC-binding peptides of flexible lengths from sequence-derived structural and physicochemical properties.

Peptide binding to MHC is critical for antigen recognition by T-cells. To facilitate vaccine design, computational methods have been developed for predicting MHC-binding peptides, which achieve impressive prediction accuracies of 70-90% for binders and 40-80% for non-binders. These methods have been developed for peptides of fixed lengths, for a limited number of alleles, trained from small number of non-binders, and in some cases based straightforwardly on sequence. These limit prediction coverage and accuracy particularly for non-binders. It is desirable to explore methods that predict binders of flexible lengths from sequence-derived physicochemical properties and trained from diverse sets of non-binders. This work explores support vector machines (SVM) as such a method for developing prediction systems of 18 MHC class I and 12 class II alleles by using 4208-3252 binders and 234,333-168,793 non-binders, and evaluated by an independent set of 545-476 binders and 110,564-84,430 non-binders. Binder accuracies are 86-99% for 25 and 70-80% for 5 alleles, non-binder accuracies are 96-99% for 30 alleles. Binder accuracies are comparable and non-binder accuracies substantially improved against other results. Our method correctly predicts 73.3% of the 15 newly-published epitopes in the last 4 months of 2005. Of the 251 recently-published HLA-A*0201 non-epitopes predicted as binders by other methods, 63 are predicted as binders by our method. Screening of HIV-1 genome shows that, compared to other methods, a comparable percentage (75-100%) of its known epitopes is correctly predicted, while a lower percentage (0.01-5% for 24 and 5-8% for 6 alleles) of its constituent peptides are predicted as binders. Our software can be accessed at .

Prediction of factor Xa inhibitors by machine learning methods.

Factor Xa (FXa) inhibitors have been explored as anticoagulants for treatment and prevention of thrombotic diseases. Molecular docking, pharmacophore, quantitative structure-activity relationships, and support vector machines (SVM) have been used for computer prediction of FXa inhibitors. These methods achieve promising prediction accuracies of 69-80% for FXa inhibitors and 85-99% for non-inhibitors. Prediction performance, particularly for inhibitors, may be further improved by exploring methods applicable to more diverse range of compounds and by using more appropriate set of molecular descriptors. We tested the capability of several machine learning methods (C4.5 decision tree, k-nearest neighbor, probabilistic neural network, and support vector machine) by using a much more diverse set of 1098 compounds (360 inhibitors and 738 non-inhibitors) than those in other studies. A feature selection method was used for selecting molecular descriptors appropriate for distinguishing FXa inhibitors and non-inhibitors. The prediction accuracies of these methods are 89.1-97.5% for FXa inhibitors and 92.3-98.1% for non-inhibitors. In particular, compared to other studies, support vector machine gives a substantially improved accuracy of 94.6% for FXa non-inhibitors and maintains a comparable accuracy of 98.1% for inhibitors, based-on a more rigorous test with more diverse range of compounds. Our study suggests that machine learning methods such as SVM are useful for facilitating the prediction of FXa inhibitors.

Machine learning approaches for predicting compounds that interact with therapeutic and ADMET related proteins.

Computational methods for predicting compounds of specific pharmacodynamic and ADMET (absorption, distribution, metabolism, excretion and toxicity) property are useful for facilitating drug discovery and evaluation. Recently, machine learning methods such as neural networks and support vector machines have been explored for predicting inhibitors, antagonists, blockers, agonists, activators and substrates of proteins related to specific therapeutic and ADMET property. These methods are particularly useful for compounds of diverse structures to complement QSAR methods, and for cases of unavailable receptor 3D structure to complement structure-based methods. A number of studies have demonstrated the potential of these methods for predicting such compounds as substrates of P-glycoprotein and cytochrome P450 CYP isoenzymes, inhibitors of protein kinases and CYP isoenzymes, and agonists of serotonin receptor and estrogen receptor. This article is intended to review the strategies, current progresses and underlying difficulties in using machine learning methods for predicting these protein binders and as potential virtual screening tools. Algorithms for proper representation of the structural and physicochemical properties of compounds are also evaluated.