Prediction of complications in diabetes mellitus using machine learning models with transplanted topic model features

Purpose: This study aims to predict the progression of Diabetes Mellitus (DM) from the clinical notes through machine learning based on latent Dirichlet allocation (LDA) topic modeling. Particularly, 174,427 clinical notes of DM patients were collected from the electronic medical record (EMR) system of the Seoul National University Hospital outpatient clinic. Method: We developed a model to predict the development of DM complications. Topics developed by the topic model were exploited as the key feature of our machine-learning model. The proposed model generalized a correlation between topic structures and complications. Results: The model provided acceptable predictive performance for all four types of complications (diabetic retinopathy, diabetic nephropathy, nonalcoholic fatty liver disease, and cerebrovascular accident). Upon employing extreme gradient boosting (XGBoost), we obtained the F1 scores of the predictions for each complication type as 0.844, 0.921, 0.831, and 0.762. Conclusion: This study shows that a machine learning project based on topic modeling can effectively predict the progress of a disease. Furthermore, a unique way of topic model transplanting, which matches the dimension of the topic structures of the two data sets, is presented.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13534-023-00322-7.

[Preliminary experience of surgical treatment for torus tubarius hypertrophy in children]

Objective: To assess the incidence of symptomatic torus tubarius hypertrophy (TTH) in recurred OSA in children, and to explore the preliminary experience of partial resection of TTH assisted with radiofrequency ablation. Methods: From January 2004 to February 2020, 4 922 children, who diagnosed as OSA and received adenotonsillectomy at the Department of Otolaryngology, The 4th Medical Center of the PLA General Hospital, were retrospectively reviewed. There were 3 266 males and 1 656 females, the age ranged from 1 to 14 years old(median age of 5.0 years). Twenty-two cases were identified with recurrence of OSA syndrome, and the clinical data, including sex, age of primary operation, age of recurrence and presentation, and opertation methods were analyzed. Follow-up was carried out by outpatient visit or telephone. Graphpad prism 5.0 software was used for statistical analysis. Results: Twenty-two cases were identified as recurred OSA and received revised surgery in 4 922 cases. Among these 22 cases, 11 cases were diagnosed as TTH resulting in an incidence of 2.23‰(11/4 922), 1 case was cicatricial adhesion on tubal torus (0.20‰, 1/4 922), 10 cases were residual adenoid combined with tubal tonsil hypertrophy (2.03‰, 10/4 922). Median age of primary operation was 3.0 years (range:2.4 to 6.0 years) in 11 TTH cases. Recurrent interval varied from 2 months to 5.5 years (2.4±1.9 years) after first operation. Age of revised partial resection of TTH was 7.0±2.7 years (range: 4.0 to 12.0 years). Average time interval between primary operation and revised operation was 3.5±2.1 years (range: 0.5 to 6.0 years). Individualized treatments were carried out based on partial resection of TTH assisted with radiofrequency ablation. All of 11 cases received satisfied therapeutic results without nasopharyngeal stenosis occured. Twenty-two cases were followed up for 1.6 to 13 years (median follow-up time was 6.2 years). Conclusions: TTH contributed to recurred OSA in child. TTH might be misdiagnosed as tubal tonsil hypertrophy. Partial resection of TTH assisted with radiofrequency ablation was a safty and effective treatment.

An integrated mathematical model of thrombin-, histamine-and VEGF-mediated signalling in endothelial permeability

BACKGROUND

Endothelial permeability is involved in injury, inflammation, diabetes and cancer. It is partly regulated by the thrombin-, histamine-, and VEGF-mediated myosin-light-chain (MLC) activation pathways. While these pathways have been investigated, questions such as temporal effects and the dynamics of multi-mediator regulation remain to be fully studied. Mathematical modeling of these pathways facilitates such studies. Based on the published ordinary differential equation models of the pathway components, we developed an integrated model of thrombin-, histamine-, and VEGF-mediated MLC activation pathways.

RESULTS

Our model was validated against experimental data for calcium release and thrombin-, histamine-, and VEGF-mediated MLC activation. The simulated effects of PAR-1, Rho GTPase, ROCK, VEGF and VEGFR2 over-expression on MLC activation, and the collective modulation by thrombin and histamine are consistent with experimental findings. Our model was used to predict enhanced MLC activation by CPI-17 over-expression and by synergistic action of thrombin and VEGF at low mediator levels. These may have impact in endothelial permeability and metastasis in cancer patients with blood coagulation.

CONCLUSION

Our model was validated against a number of experimental findings and the observed synergistic effects of low concentrations of thrombin and histamine in mediating the activation of MLC. It can be used to predict the effects of altered pathway components, collective actions of multiple mediators and the potential impact to various diseases. Similar to the published models of other pathways, our model can potentially be used to identify important disease genes through sensitivity analysis of signalling components.

Identifying Novel Type ZBGs and Nonhydroxamate HDAC Inhibitors Through a SVM Based Virtual Screening Approach

Histone deacetylase inhibitors (HDACi) have been successfully used for the treatment of cancers and other diseases. Search for novel type ZBGs and development of non-hydroxamate HDACi has become a focus in current research. To complement this, it is desirable to explore a virtual screening (VS) tool capable of identifying different types of potential inhibitors from large compound libraries with high yields and low false-hit rates similar to HTS. This work explored the use of support vector machines (SVM) combined with our newly developed putative non-inhibitor generation method as such a tool. SVM trained by 702 pre-2008 hydroxamate HDACi and 64334 putative non-HDACi showed good yields and low false-hit rates in cross-validation test and independent test using 220 diverse types of HDACi reported since 2008. The SVM hit rates in scanning 13.56 M PubChem and 168K MDDR compounds are comparable to HTS rates. Further structural analysis of SVM virtual hits suggests its potential for identification of non-hydroxamate HDACi. From this analysis, a series of novel ZBG and cap groups were proposed for HDACi design.

Trace metals in different species of mollusca, water and sediments from Taiwan coastal area

Since October 1994, a long-term program of Asia/Pacific Mussel Watch: Taiwan Regional Studies has been carried out. The results indicate that trace metal contents in mollusca varied among 30 different species and the environments (water and sediments) along the Taiwan coast. The orders of bioaccumulation of trace metals in mollusca were: Cu (over 200 microg/g), Thais clavigera > Isognomon legumen > Clibanarius rivescens > Crassostrea gigas; Zn (over 700 microg/g), Moruta granulata > C. gigas > Asiatica cypraea arabica > T. clavigera; Cd (over 5 microg/g), Trochus hanleyanus > Acanthopleura japonica > Nerita albicilla > Patella flexuosa; Pb (over 2 microg/g), P. flexuosa > C. gigas > T. hanleyanus > T. clavigera > C. gigas > Capiyulum mitella; Ni (over 10 microg/g), Meretrix lusoria > Philine sp. > Littoraria scabra > Tridacna squamosa > T. hanleyanus; Cr (over 30 microg/g), Littoraria undulata > T. hanleyanus > N. albicilla > Nerita chamaelor > M. granulata; As (over 20 microg/g), Perna viridis > L. scabra; and Sn (over 5 microg/g), P. viridis > L. undulata> C. mitella> C. gigas. Their seasonal and regional variations as well as their correlation are evaluated and discussed.

Estimation of potential lifetime cancer risks for trihalomethanes from consuming chlorinated drinking water in Taiwan

Data on concentrations of trihalomethanes (THMs) in raw and chlorinated water collected from three water treatment plants in Taiwan and estimates of the lifetime cancer risk for THMs from drinking water, using age-adjusted factors and volatilization terms, are presented. Data on THM levels in drinking water were obtained from the annual reports of the Environmental Protection Administration (EPA) of Taiwan. The methodology for estimation of lifetime cancer risks was taken from the USEPA. Chloroform was the major species of THMs, especially in the water plant of south Taiwan. Chloroform contributed the majority of the lifetime cancer risks (range: 87.5-92.5%) of total risks from the three water supply areas. All lifetime cancer risks for CHCl(3), CHBrCl(2), CHBr2Cl, and CHBr3 from consuming tap water in the three water supply areas were higher than 10(-6). The sum of lifetime cancer risks for CHCl(3), CHBrCl(3), CHBr2Cl, and CHBr3 was highest (total risk for total THMs<1.94x10(-4)) for tap water from south Taiwan.

Mussel watch: a review of Cu and other metals in various marine organisms in Taiwan, 1991-98

This study presents the distribution of Cu, Zn, Pb, Cd, Hg and As in various marine organisms collected along the western coast of Taiwan from 1991 to 1998, and also evaluates the time variation of Cu in oysters before (1980-85) and after (1986-98) the "green oyster" incident. The results show that relatively high geometric mean (GM) concentrations of Cu, Zn, Pb, Cd, As and Hg were generally found in Crassostrea gigas (Cu=229 microg/g, Zn=783 microg/g), Gomphina aeguialtera (Pb=30.3 microg/g), Tegillarca granosa (Cd=2.85 microg/g), Thais clavigera (As=96.9 microg/g) and Parapenaeopsis cornuta (Hg=1.35 microg/g), respectively. Especially, maximum Cu and Zn concentrations (GM=229 and 783 microg/g, respectively) in oysters (C. gigas) from different culture areas were much higher than those of the other organisms by about 1.13-458 and 2.40-63.7 times, respectively. Similarly, rock-shells (Thais clavigera) had a high capacity for accumulating Cu (GM=202 microg/g) and Zn (GM=326 microg/g) under the same physico-chemical conditions. The highest GM Cu and Zn concentrations of 1108 (range from 113 to 2806) and 1567 (range from 303 to 3593) microg/g were obtained in oysters from the Hsiangshan coastal area, one of the most important oyster culture areas in Taiwan. However, the highest GM Cd and As concentrations of 6.82 and 19.3 microg/g were found in oysters from the Machu Islands. Mean Cu concentrations in the oysters from the Erhjin Chi estuary declined from 2194+/-212 microg/g in 1986-90 to 545 microg/g (GM) in 1991-96. In the Hsiganshan area, GM Cu concentrations of 909 microg/g (1991-96) and 1351 microg/g (1997-98) in oysters were significantly higher than those of 201 microg/g (1980-85) and 682 microg/g (1986-90). The gradually increasing levels of Cu and Zn in the oysters from the Hsiangshan area have been observed year by year.

Estimation of metal and organochlorine pesticide exposures and potential health threat by consumption of oysters in Taiwan

Pollutant concentrations detected in oysters from 12 different culture areas of Taiwan (especially for the Hsiangshan area and the Machu Islands) from 1991-98 were evaluated to investigate potential carcinogenic (inorganic As and organochlorine pesticides) and non-carcinogenic (Cu, Zn, Cd and inorganic As) risk to the public from ingestion of the oysters. The highest geometric mean (GM) Cu and Zn concentrations of 1108 (range 113-2806) and 1567 (range 303-3593) microg/g dry weight were obtained in oysters from the Hsiangshan coastal area. The maximum GM Cd and As concentrations of 6.82 and 19.3 microg/g dry weight were found in oysters from the Machu Islands area. The p,p'-DDE values range from not detectable in Penghu Islands' oysters to 164 ng/g dry weight in Machu Islands' oysters. The highest tDDT (sum of p,p'-DDE, p,p'-DDD and p,p'-DDT) concentrations of 337 and 340 ng/g dry weight were found in oysters from Kimmen and Machu Islands, respectively. A calculated target hazard quotient (THQ; daily intake/reference dose) of 11.4 (based on 139 g oysters/day) for Cu caused by consuming oysters from the Hsiangshan area is higher than that from other areas (range 0.124-5.95). The highest average Cu intake from Hsiangshan's oysters for individuals is 11.4 times (i.e. THQ=11.4) more than that of reference dose (40 microg/kg/day). However, the maximum THQ values for Cd and As caused by consuming oysters collected from the Machu Islands were 5.57 and 2.63 for Cd and As, respectively. Generally, the results of THQ showed that if only the maximally exposed individuals were considered, the value of 65.4% for oyster was higher than 1.0 in comparison with reference dose. All cancer risk estimates for inorganic As from consuming oysters were higher than 10(-6) (range from 128x10(-6) to 509x10(-6) for maximally exposed individuals and range from 17.1x10(-6) to 68.0x10(-6) for typically exposed individuals, respectively); that is the risk of the lower end of the range of acceptable risk. The highest risk estimate for inorganic As was 509x10(-6) for consumption of oysters by Machu Islands' residents. The lifetime cancer risks of 19.0x10(-6) for tDDT by consuming oysters from the Machu Islands was higher than those from the Penghu Islands (0.37x10(-6)). Therefore, the sum of lifetime cancer risks for tDDT and inorganic As had the highest risks (total risk=528x10(-6)) of consuming oysters from the Machu Islands. Furthermore, a 10(-6) upper limit on lifetime risk as the health protection standard would require maximum oyster consumption rates of approximately 0.26 g/day.

Rock-shells (Thais clavigera) as an indicator of As, Cu, and Zn contamination on the Putai Coast of the black-foot disease area in Taiwan

This study presents the distribution of arsenic (As), copper (Cu), and zinc (Zn) in various seafoods-oysters (Crassostrea gigas), false fusus (Hemifuscus tuba), venus clams (Cyclina sineasis), green mussels (Perna viridis), blood clams (Arca granosa), flounders (Psettodes erumei), and rock-shells (Thais clavigera) collected from the Putai coast of the black-foot disease (BFD) area in Taiwan. Special attention is paid to evaluate the relationships among As, Cu, and Zn and effect of body size on metal concentration in Thais clavigera. Maximum Zn and Cu geometric mean (GM) concentrations (GM = 615 and 376 microg/g, dry wt, respectively) are found in oysters (Crassostrea gigas), and the values are much higher than those of the other organisms by about 1.7-208 and 1.8-375 times, respectively. Similarly, Thais clavigera has a high capacity for accumulating Cu and Zn collected from the same location. One interesting point is that relatively high As concentrations (GM = 65.7 microg/g, dry wt) in Thais clavigera are found as compared with those in other organisms (range from GM = 2.37 to 40.2 microg/g, dry wt). The As concentrations are significantly higher in Thais clavigera (1.62-27.7 times) than those in other organisms (p < 0.05), except for the false fusus (Hamifuscus tuba). A linear regression analysis shows a significant increase in Zn concentration with increasing Cu concentration in Thais clavigera. On the other hand, the As concentration is correlated with Cu and Zn concentrations (r = 0.77 and 0.77, respectively; p < 0.05) in Thais clavigera. Double logarithmic plots of metal content and concentration against dry-body weight and shell length show linear relationships. The result indicates that large individuals have higher contents of Cu, Zn and As, and have slopes of 1.58, 1.38, and 1.34, respectively. In addition, metal concentrations against shell length for all animal sizes also indicate that Cu, Zn and As have slopes of 1.92, 1.18, and 1.11, respectively. In conclusion, Thais clavigera has a high capactiy for accumulating As, Cu, and Zn and is a potential bioindicator for monitoring As, Cu and Zn.

Monitoring lead pollution near a storage battery recycling plant in Taiwan, Republic of China

This study presents the distribution of blood lead levels and lead in various environmental samples (water, sediments, soils, and air) near the Shing-Yie storage battery recycling plant in Taiwan before (July 1990 to June 1991) and after (July 1992 to June 1993) amelioration. Before amelioration, the average blood lead levels in the neighborhood of the plant were in the range of 10.55 +/- 5.7 to 12.28 +/- 7.9 micrograms/dl. After amelioration, relatively lower average concentrations of blood lead (range 8.35 +/- 3.0 to 9.13 +/- 2.5 micrograms/dl) were generally found; however, these averages were still higher than that (7.79 +/- 3.5 micrograms/dl) from other lead-unpolluted areas of Taiwan. An exceedingly high geometric mean (GM) lead concentration (128 micrograms/L) was found in the downstream river water of the Tawulum River passing by the plant. The concentrations of lead (GM = 372 and 418 micrograms/g) in the downstream river sediments were higher than those (GM = 123 and 158 micrograms/g) in the upstream river sediments before and after amelioration, respectively. Furthermore, lead species in river sediments were analyzed by a sequential leaching technique. The sum of phases I, II, and III accounted for 83.7% of total lead at station R2 (nearest to the plant). Maximum lead concentration (GM = 2402 micrograms/g) in dust at the soil surface from station S1 (nearest to the plant) was much higher than those from the other stations by about 18 times before amelioration. However, the maximum value dropped to 1,155 micrograms/g after amelioration. On the whole, the geometric mean concentration of lead in dust at the soil surface nearest to the plant was > 1,000 micrograms/g and decreased to < 100 micrograms/g in the 15-30 cm depth soil about 2 km away from the plant. Before amelioration, the geometric mean lead concentration of 4.57 micrograms/m3 (range 0.102-37.6 micrograms/m3) in the air near the plant was higher than that at the background locations, the geometric mean value of which was 0.08 micrograms/m3.

Relationship between copper speciation in sediments and bioaccumulation by marine bivalves of Taiwan

This paper evaluates the relationships between copper species in sediments and accumulation by the purple clam (Hiatula diphos) and venus clam (Gomphina aeguilatera) collected from the field and culture (aquaculture) ponds in the polluted coastal area of Lukang, Taiwan. Sediment was sampled along with the molluscs, including oysters (Crassostrea gigas), purple clams (Hiatula diphos), rock-shells (Thais clavigera), venus clams (Gomphina aeguilatera), and hard clams (Meretrix lusoria), from two unique environments of Lukang during the period from August 1993 to July 1994. The data indicate that the total copper concentrations in sediments from culture ponds (185 microg g(-1)) was higher than those of the field (44.0 microg g(-1)). Copper species in sediments were analyzed by a sequential leaching technique. Results show that concentrations of various copper species in the sediments are in the range of 1.14 +/- 0.59 to 13.2 +/- 22.4 microg g(-1) and 0.36 +/- 0.24 to 133 +/- 36.7 microg g(-1) for the two environments, respectively. Also the exchangeable copper in sediment from culture ponds was 15 times higher than that from the field. In addition, the sum of exchangeable and copper carbonates had the highest percentages of copper in both the pond sediment (86.6 %) and the field sediment (50.7 %). Maximum copper concentrations (309 +/- 35.1 microg g(-1)) in oysters were much higher than those in the other benthic organisms by about 4-127 times. Similarly, the data also showed that copper concentrations in Thais clavigera were 12-32 times higher than those in other benthic organisms. Copper concentrations in various benthic organisms differed significantly (p < 0.05) from that in Thais clavigera. This capacity makes Thais clavigera a potential candidate for monitoring copper in marine sediments. In terms of copper species, the best correlation was generally obtained between copper carbonates in sediments and copper concentrations in Hiatula diphos (r = 0.886*). A strong multiple regression correlation (p < 0.05, r2 = 0.7894) also indicates that the copper carbonates may dominate as the available form of copper to Hiatula diphos from various environments in the Lukang coastal area under natural physicochemical conditions.

Coprostanol distribution in marine sediments off southwestern Taiwan

One of the major industries in southern Taiwan is pigfarming along the Kaoping River; some two million animals are being raised along the river banks. Excretions from pigs, treated and untreated, are discharged directly into the river and eventually carried to the sea. Twenty-four surface sediments and one sediment core off southwestern Taiwan were analyzed to determine quantitatively the extent of coprostanol addition and its distribution and to obtain the input of coprostanol over the past. Geographically, the percent coprostanol is highest around the river mouth, and decreases to about 1% at the shelf break and about 0% at a distance of approximately 40 nautical miles (74 km) from the river mouth. The progressive seaward decline of percent coprostanol from the river mouth can be attributed to [1] dilution of coprostanol by uncontaminated sediment and/or sediment containing relatively lower levels of coprostanol, [2] dilution of coprostanol by biogenic sterols, and [3] probably degradation of coprostanol. Further, the Kaoping Canyon sediments contain relatively high percent coprostanol; this can be attributed to [1] a more direct input of the river sediments because the canyon is well aligned with the river, [2] currents in the canyon being alternate upcanyon and downcanyon which tend to keep sediments in the canyon, and [3] the possible blocking effect of a topographic high in the canyon. A sediment core exhibits comparatively higher percent coprostanol in the top 15 cm, indicating an increased input of coprostanol over the past 20 years.

Transient and sustained components of the pupillary responses evoked by luminance and color

That the pupil reacts to changes in luminance and color, as well as to spatial features in the retinal image raises questions about whether phasic and tonic and/or color and luminance visual pathways project to the pretectal pupillomotor neurons. The present study compares pupillary responses evoked by heterochromatic and achromatic luminance increments to investigate whether the pupillary responses evoked by color and by luminance are independent of one another. Principal component analysis is used to examine the constituents of the pupil responses. The results support the belief that the visual input to the pupillomotor system is organized into phasic and tonic (but not necessarily independent color and luminance) pathways.

Depuration of copper and zinc by green oysters and blue mussels of Taiwan

This paper describes depuration processes of copper and zinc in green oysters (Crassostrea gigas) and in blue mussels (Mytilus smarangdium) collected from an environment with heavy copper contamination, and then transferred to natural clean seawater. Results show that the total loss of copper content per oyster is an exponential function of exposure time for the first 6 days with a depuration rate of 351 microg g(-1) day(-1) and then levels off. During this exponential decrease period approximately 67% of the copper accumulated in green oysters was depurated. However, when the copper contents in the oysters decreased from 2225 +/- 111 microg g(-1) to 344 +/- 18.7 microg g(-1) the depuration rates decreased from 245 microg g(-1) day(-1) to 0.08 microg g(-1) day(-1). This means that green oysters had a 16-fold higher copper depuration rate (351 microg g(-1) day(-1)) than normal oysters (21.5 microg g(-1) day(-1)) for the first 6 days. However, the depuration of accumulated copper and zinc by the mussels was a fast process in natural clean seawater. About 91% of the accumulated copper was lost during the first 6-day period; copper contents declined from 20.2 +/- 3.41 microg g(-1) to 1.80 +/- 0.21 microg g(-1). Only 36% of the accumulated zinc was lost during a depuration period of 6 days. Calculations show that the biological half-lives of copper in green and normal oysters were 11.6 and 25.1 days, respectively. The biological half-lives of zinc in green and normal oysters were 16.7 and 30.1 days, respectively. In spite of the relatively low initial copper content in blue mussels being 20.2 +/- 3.41 microg g(-1), the biological half-life is only 6.40 days. From these results it is important to emphasise that the fastest turnover rate is for copper in blue mussels. However, zinc is more retentive in blue mussels than copper.

Green oysters caused by copper pollution on the Taiwan coast

The first case of green oysters (Crassostrea gigas) broke out along the Charting mariculture area of south-western Taiwan in January 1986. The green color was found to be due to high copper content in the oyster tissue. Since then, a long-term survey around this area shows that total dissolved copper ranges from 4.99 to 23.6 microg/liter and particulate copper ranges from 1.09 to 5.51 microg/liter in sea-water. The green oysters collected from the Erhjin Chi estuary on 26 January, 1989 gave the highest copper content, 4401+/-79 ppm dry wt. Other green oyster cases were occasionally observed in the Hsiangsan and Anpin mariculture areas. Meanwhile, an experiment of copper accumulation in oysters was conducted at three stations (south-western Taiwan) for up to 90 days. Multiple regression analysis indicates that the food pathway may predominate in copper accumulation by green oysters. This bioaccumulation experiment shows that the total uptake of copper per oyster is an exponential function of exposure time for the first 2 weeks with an accumulation rate of 214 ppm Cu/day and then levels off. The average values of concentration factors for oysters (about 5 x 10(5)) were very close to steady-state values under the natural conditions at each station.