The Risk Assessment of Thyroid Disorders Based on Metabonomics

50 healthy volunteers and 200 patients with thyroid diseases (58 cases of thyroid nodule, 59 cases of goiter, 34 cases of postoperative thyroid cancer and 24 cases of thyroid cancer) were compared in our study. Total trace elements of iodine, selenium, manganese, chromium etc. (in total 28 elements) in human serum were detected by inductively coupled plasma mass spectrometry (ICP-MS). A risk assessment model for thyroid disease was established after analyzing trace elements content, thyroid function and ultrasonography results by gender and age with the characteristics of the Chinese population. Ten repeat collections for urinary iodine from spot samples or 24-hour samples are needed to reliably estimate individual iodine status in women.

cancers according to previous research [3]. Worldwide, the incidence rate of thyroid disease including thyroid cancer was increasing. In China, thyroid cancer ranks third female solid tumors, and has become the fastest-growing incidence of malignant solid tumors [4].
Because the symptoms of thyroid diseases are hidden, the high-risk group should check the thyroid function according to their symptoms regularly. The current examines include high-resolution ultrasound, hypersensitive thyroid stimulating hormone (TSH), ne needle aspiration biopsy (FNAB), and analysis combined with clinical history and physical examination ndings. However, the pathological reports of patients who cannot diagnose the nature of thyroid nodules before operation show that 20% of the thyroid nodules are malignant and 80% of the patients receive unnecessary partial or total thyroidectomy [5]. Once the surgery is operated, the life quality of patients will be seriously affected. Unaesthetic and patients need hormone replacement therapy for life after surgery. On the other hand, the medical costs involved are also signi cantly different [6] which seriously consumes medical and social resources, and has become an important public health problem in China. which can also be called as elemental omics [7][8]. Trace elements play an important role in the metabolism and growth of the human body. Thyroid disease will affect the metabolism balance of trace elements in the body. In turn, the imbalance of trace elements will aggravate the condition of patients with thyroid disease. It is reported that the level of urinary iodine is positively correlated with the incidence rate of thyroid cancer, and there is a signi cant correlation between high urinary iodine and thyroid malignancy and thyroid cancer metastasis. Selenium level is negatively correlated with thyroid cancer incidence rate. [9][10] At present, there are few reports about the relationship between metabonomics and clinical diseases. This paper studies the relationship between trace elements and thyroid diseases, to provide new ideas for the treatment of thyroid diseases.
With the increasingly clear relationship between trace elements and diseases, accurate, rapid, and convenient determination of trace elements in the human body has become an urgent issue in the eld of trace element detection. Therefore, the detection of trace elements, early screening of thyroid diseases, treatment, and prognosis judgment based on personal characteristics of drugs and treatment, to provide a scienti c basis for clinical decision-making, can effectively reduce personal pain and social medical burden.
On the basis of detecting the element group of normal people and patients with thyroid disease, this paper attempts to establish a Random forest model for risk assessment of thyroid nodule with iodine, selenium, calcium, iron, zinc and other trace elements, providing scienti c and effective data support for clinical diagnosis and treatment of thyroid nodule, to promote standardized diagnosis and treatment of thyroid nodule in China. Meanwhile, avoiding benign nodule operation can improve the early detection rate of thyroid cancer, reduce the death risk of thyroid cancer, and improve the survival rate and life quality.

Subjects
This study was conducted in the Department of Medical Laboratory, Tongren Hospital from January 2019 to December 2019. It was performed following the guidelines of the Declaration of Helsinki Principles and approved by the Institutional Review Board of Tongren Hospital (approval number 2019-0907). All subjects gave written informed consent. 50 healthy people with physical examination were randomly selected as the control group from January to December 2019. Above mentioned 50 healthy people who received a comprehensive examination in the physical examination center of our hospital were healthy people without an adenoid disease (normal thyroid).
The enrollment criteria were: 200 cases of thyroid operation in Tongren Hospital, Capital Medical University, older than 18 years old with pathological results con rmed thyroid cancer, thyroid nodule diameter is not less than 1cm. Among them, 58 cases of thyroid nodule, 59 cases of goiter, 34 cases of postoperative thyroid, and 24 cases of thyroid cancer, other types of patients cannot be statistically analyzed due to the small sample size (less than 3 cases). Exclude Standard: thyroid function test and ultrasonography results were not from our hospital (ensure the measurement method and level of data sources are consistent); combination a serious chronic disease, such as kidney failure, liver disease, or heart disease; cancer; recurrent TC; thyroid autoimmune disease; surgical history within 6 months before the study; smoking; take thyroid related drugs before testing; pregnantor lactating women. Sampling 10 mL venous blood was collected, coagulated for 30 min, then centrifuged for 10 min (16,000 g, 4°C). Upper phase of serum was taken and stored in 2 mL Eppendorf tubes at -80℃ before analysis.

Analysis
Thyroid volume was measured by using GE LOGIQ 9 and Philips iu22 colors Doppler ultrasound diagnostic instrument and linear array probe (frequency 5-14 MHz). The concentrations of serum FT3, FT4, T3, T4, TSH, TgAbs, TPOAb, and TSH were measured by Chemiluminescent immunoassay analysis using reagent pack (Roche Germany). The following are reference ranges: 3.1-6.8 pmol/L for FT3, 12-22 pmol/L for FT4, 1.3-3.1 nmol/L for T3, 66-181 nmol/L for T4, 0.27-4.2 mIU/L for TSH, <115 IU/mL for TgAbs, and 0-34 IU/mL for TPOAb. ICPMS (Clin-ICP-QMS-I, Bioyong Technology Co., Ltd., Beijing, China) was used for trace element analysis. The content of trace elements in the serum is very low (ng/L~μg/L) and there are several uncertain factors in the detection process caused by the complex matrix of human serum. The collision reaction cell is helpful to reduce interference and increase the detection limit of elements. Therefore, we use He mode to detect all 28 elements in this test. The basic parameter settings of the instrument are shown in Table 1.
Ultrapure water used in the experiment was prepared by Milli Q ultrapure water meter (Millipore, Austria), Nitric acid (Rotipuran ® ≥ 65%, P.A., ISO, Roth, Karlsruhe, Germany) puri ed by low boiling point distillation. The single standard solution used in the experiment was calcium, iron, zinc, iodine, lithium, vanadium, chromium, manganese, cobalt, nickel, copper, gallium, arsenic, selenium, strontium, strontium, molybdenum, palladium, cadmium, tin, antimony, tellurium, barium, cerium, platinum, mercury, thallium, lead, bismuth, germanium, germanium, indium, terbium, and rhodium. (1000 μg/ml, National Center for The blank reagent, standard curve, and quality control used in the measurement were prepared as the same matrix. The polypropylene tubes used in the sample preparation process were soaked overnight with 10% nitric acid and washed with ultrapure water for three times. The standard curve was prepared by mixing a single standard solution. For the six-point standard curve of serum iodine, 5% (V/V) ammonia solution was diluted to a concentration of 60, 80, 120, 160, 180, and 200 μg/L; for the internal standard, indium was selected, and the internal standard element was also diluted to a nal concentration of 20 μg/L by 5% (V/V) ammonia solution. For the six-point standard curve of serum calcium, 5% (V/V) ammonia solution is used to dilute to 2, 5, 10, 50, 100, and 200 μg/L; for the six-point standard curve of other elements in serum, 2% (V/V) nitric acid solution is used to dilute to 0, 40, 80, 120, 160, and 200 μg/L; for the internal standard, germanium, indium, rhodium, and terbium were selected; for the internal standard element, 5% (V/V) nitric acid solution was also used to dilute to the nal concentration of 20 μg/L.

Statistical Analysis
The software of Clin-ICP-QMS-I was used for quantitative analysis, and Excel (Microsoft Corp., Redmond, WA) and OriginPro9.0 (Northampton, Ma, USA) were used for data statistical analysis. Python 3.6.4 was used for the random forest model. The experimental data were expressed as median and quartiles due to their nonnormal distribution. The correlation between experimental variables was evaluated using Spearman's rank correlation coe cients. Mann-Whitney U test was implemented for non-parametric variables. Different logistic regression models were used to assess the relative risk of a thyroid nodule. All tests were repeated three times and considered statistically signi cant at p< 0.05. Results 1.1 Comparison between healthy people (n=50) and patients with thyroid disease (n=200) Comparison between healthy people (n=50) and patients with thyroid disease (n=200) are shown in Table  2 In table 2, there were eight elements showed signi cant difference between thyroid disease and control group. Except iodine, the content of all the other elements are relatively low in the serum of thyroid patients compare to the control group. While the content of molybdenum, cadmium, tin, palladium, and thallium in the healthy people and thyroid disease patients is not signi cantly different.

Comparison between healthy people and patients with thyroid disease by gender
A comparison of the classi cation of healthy people and patients with thyroid diseases by gender were shown in Table 3. The results of females were shown in Table 4. The results of males were shown in Table 5.
Signi cant differences in ten trace elements of iodine, calcium, chromium, manganese, cobalt, nickel, copper, zinc, selenium, and antimony were found between female healthy people and female patients with thyroid disease (P<0.05). The content of iodine, chromium, and manganese were increased, while the content of the rest elements were decreased.
Signi cant differences in the eight trace elements of iodine, manganese, iron, nickel, copper, zinc, selenium, and antimony were found between the male healthy people and male patients with thyroid diseases (P<0.05). Except iodine and manganese, the content of rest elements were decreased in the patients with thyroid diseases compare to the control group.

Correlation between trace elements
We assessed the correlation between concentrations of essential trace elements in both the thyroid disease group and control group. The correlation coe cient is listed in Table 6, and the con dence ellipse between trace elements in thyroid group were shown in Figure 1.
In the thyroid disease group, there were strong correlations between Fe-Zn, Cr-Mn, Ca-Zn, Ca-Se, Ca-Fe and Zn-Se. Whereas, in the control group, the strong correlation only found between Ca-Zn and Ca-Se, no such correlations were found between other elements. Compared with the control group, the correlation between Ca-Zn and Ca-Se in the thyroid disease group were stronger. (Figure 1).
1.6 Establishment of thyroid disease risk assessment model and its effect evaluation Since the accuracy of regression analysis of thyroid diseases (thyroid cancer, thyroid nodule, goiter, and postoperative thyroid surgery) and the classi cation of healthy people is not very high, we use a random forest algorithm for effective classi cation.
The values of each index in the data are not all in an order of magnitude. We use the normalization method to preprocess the data. We use a random forest algorithm to model the following four groups of data and divide each group of data into a correction set (70%) and prediction set (30%). Because of the randomness of data set partition, each model is run 50 times and the result of prediction set is taken as the nal prediction result of each group. From the results in Table 7, we can see that random forest has a good diagnostic effect, with speci city and sensitivity of more than 95%.

Discussion
China used to be one of the countries with serious iodine de ciency in the world. Iodine de ciency may lead to abortion, stillbirth, congenital malformation, endemic cretinism, hypothyroidism, goiter, mental retardation, and other diseases, seriously affecting the quality of the population. [11][12]. Since the implementation of universal salt iodization (USI) in 1995, the research results show that the Median Urinary Iodine MUI) of 8 to 10 years old children has increased from 164.8μg/L in 1995 to 238.6μg/L in 2011, which exceeds the World Health Organization's best standard for urinary iodine (100-199μg/L), while it is at the level of iodine excess, 200-299 μg/L [13][14]. Along with the change of iodine nutritional status, the incidence rate of thyroid diseases is changing. It shows that the thyroid disease is developing from simple goiter to toxic nodular goiter, Hashimoto's thyroiditis, and thyroid cancer. [15] Among them, the incidence of thyroid malignancies which is most harmful to human beings annually increase of 14.51%, [16] and is believed to be caused by excessive iodine intake. However, this conclusion has been controversial [17]. This study shows that a high concentration of iodine is one of the risk factors of thyroid disease. Therefore, we should supplement iodine according to the content of iodine in the serum.
This study also shows that a low concentration of selenium is another of the risk factors for thyroid disease. Selenium has important biological signi cance for human beings. It is an essential trace element for the human body and has the highest selenium content in the thyroid gland. Selenoprotein is a kind of protein which active center is selenocysteine. At present, 25 kinds of selenoproteins have been found in the human body, mainly including 6 kinds: glutathione peroxidase family, iodothyronine deiodinase family , selenoprotein P, selenoprotein W, thioredoxin reductase and selenophosphate synthetase. Selenium plays a variety of biological functions in the body, such as anti-oxidation, improving immune function, anti-tumor, etc. Selenium is involved in the synthesis, activation, and metabolism of thyroid hormones, and plays an important role in thyroid antioxidant system and immune system [18][19][20]. Selenium de ciency is closely related to goiter, autoimmune thyroid disease, low T3 syndrome, thyroid cancer, and other diseases [21][22]. Selenium containing preparations can improve thyroid function in some patients with thyroid diseases, which provides a new way for the treatment of thyroid diseases [23][24][25]. A number of studies have shown that in children with goiter in low iodine area, the activities of seleniumand GPx in plasma are signi cantly reduced. It is concluded that goiter is not only related to iodine but plays an important role in low selenium. The mechanism may be as follows: low selenium can also cause metabolism disorder of tissue, especially myocardial tissue, dysfunction of mitochondrial oxidative phosphorylation, and the body needs a relatively stable T3 level to ensure that it can provide energy for normal oxidative phosphorylation; deiodinase activity is reduced, T3 production is insu cient, which is not able to meet the needs of the body's metabolism, and TSH rises to feedback to regulate more thyroid secretionT3; low selenium reduces the activity of IDI in liver and kidney, decreases the output of T3 in peripheral tissue, and increases the concentration of T4 in blood, decreases the concentration of T3; at the same time, the activity of ID in pituitary decreases, the output of T3 in pituitary decreases, the negative feedback effect of T4 decreases, and the TSH released by the pituitary increases; low selenium also affected the activity of glutathione peroxidase (GPx) in thyroid, and eliminated H 2 O 2 produced by cell metabolism The activity of TPO increased, H 2 O 2 oxidized inorganic iodine to activated iodine, accelerated the synthesis of thyroid hormone, and eliminated H 2 O 2 at the same time. Due to the above reasons, the synthesis and secretion of thyroid are enhanced, the T3, T4 and iodine reserves in thyroid tissue are insu cient, the iodine intake of thyroid is increased, the protein binding iodine is increased, and the TSH level is increased, resulting in compensatory goiter of thyroid [26][27][28].
This study shows that manganese plays an important role in thyroid disease. Manganese (Mn) is an essential trace element with potential toxicity under the high level of exposure. As a component of many enzymes and cofactors, manganese plays an important role in many physiological processes in mammals. Manganese superoxide dismutase (Mn SOD) as amanganese-containing enzymeis the main antioxidant enzyme, which can neutralize the toxic effect of active oxygen. Other manganese-containing enzymes include oxidoreductase, transferase, hydrolase, lyase, isomerase, ligase, and glutamine synthetase. Dopamine is also known as an inhibitor of TSH secretion. Dopamine and dopaminergic receptors are involved in neurodevelopment and TSH regulation. Excessive manganese exposure may lead to adverse neurodevelopment results, which is due to the loss of dopaminergic control of TSH regulation of thyroid hormones and the destruction of thyroid homeostasis. This destruction may alter thyroid hormone levels, leading to some defects related to manganese exposure during pregnancy. Dysregulation of thyroid hormone may regulate the production of dopamine through manganese, thus directly or indirectly affecting thyroid function [29].
In females with thyroid disease, calcium plays a very important role. Some studies suggest that females are more likely to have low calcium [30]. Rajinikanth et al. [31] found that the high proportion of parathyroid glands in the thyroid capsule in female patients may be the reason for the high incidence of low calcium. Besides, the gender difference may be related to the effect of sex steroids on the secretion of parathyroid hormone and the genetic variation of the cell signaling pathway. Female menstruation, pregnancy, lactation, and menopause will also cause a lack of calcium and vitamin D. There are much data to show that the lack of vitamin D involves a variety of cancer pathogenesis, including colorectal cancer, prostate cancer, pancreatic cancer, and thyroid cancer. Some experimental studies have proved that vitamin D can induce apoptosis and delay the growth of cancer cells. Clinckspoor et al [32] con rmed the inhibitory effect of 1, 25-dihydroxyvitamin D3 on the proliferation of thyroid cancer cells.
The relationship between vitamin D de ciency and thyroid diseases has become a hot research topic and most research evidence supports the view that vitamin D de ciency can increase the risk of multiple thyroid diseases. However, it is not clear whether vitamin D de ciency has some special mechanism. Furthermore, the vitamin D level of the body is affected by many factors, thus the results of many studies are still objective differences. Regardless of whether vitamin D supplementation has the effect of preventing and treating thyroid diseases, how to supplement vitamin D still needs a lot of research support.
Although there is a signi cant difference in iron between male patients with thyroid disease and male healthy control group, the incidence of iron de ciency anemia in female patients is relatively high. Many enzymes in the human body need iron as a cofactor, including enzymes involved in the tricarboxylic acid cycle and the electron transfer system. In the absence of iron, the activity of iron-containing enzymes and iron-dependent enzymes in the body is reduced [33], the biochemical reactions catalyzed by these enzymes are blocked, the metabolic function of the body is changed, and the homeostasis of the internal environment is affected. The activation, iodization, and coupling of iodine need the catalysis of TPO. Serious iron de ciency will reduce the activity of TPO, affect the response to iodine in the process of thyroid hormone synthesis, and then affect the synthesis of thyroid hormone [34]. When thyroid function is normal, T3 mainly comes from T4 to T3 transformation of liver and kidney. While, when thyroid function is low, T3 mainly comes from thyroid. The hormone T4 is transformed into T3 with biological activity through the deiodination of the outer ring. This particular pathway is catalyzed by a selenine containing type I deiodinase (ID-I), which is present in high concentrations in the liver, kidney, thyroid, and pituitary. Goswami et al. [35] found that iron ion complex 2,2' dipyridine can strongly inhibit the activity of ID-I, indicating that there may be an iron ion active center to affect the catalytic effect of the enzyme. Therefore, the activity of ID-I decreased and the 5'deiodination pathway was inhibited in the iron de ciency state, which led to the decrease of T3 in serum. It is also shown that iron de ciency anemia can reduce the response of TSH to Thyrotropin releasing hormone (TRH), and affect the regulation of the central nervous system on thyroid function [36]. Therefore, many countries advocate strengthening iodine nutrition and iron nutrition intake at the same time [37,38], which may be a better control method for IDD in iron de ciency areas.
Random forest is a kind of integrated learning model based on the decision tree, which has a relatively good classi cation effect. [39,40] It can process high-dimensional feature data, not easy to over t, and the model training speed is faster. Compared with the random forest algorithm, the logistic regression only evaluates with a single test sample, but its explanation of the in uencing factors are vague, and the relative risk of the in uencing factors cannot be given and the direction of action. However, the random forest algorithm balances the in uence of sample error by a large number of randomly selected samples and classi es and comprehensively evaluate a large number of different test data, which is more reliable than the result of logistic regression tting.
There are still some de ciencies in this study: the number of cases is not enough and a risk assessment model for patients with thyroid disease by disease type will be further established.

Conclusions
The high concentration of iodine and manganese are risk factors of thyroid disease. In addition to the low concentration of cobalt, nickel, copper, zinc, selenium, and antimony, calcium in the serum of female patients with thyroid disease and iron in the serum of male patients with thyroid disease play a very important role. Random forest model based on element risk assessment of thyroid disease is helpful for the diagnosis and treatment of thyroid disease.

Declarations
Ethics approval and consent to participate The study protocol was performed following the guidelines of the Declaration of Helsinki Principles and approved by the Institutional Review Board of Tongren Hospital (approval number 2019-0907). All participants gave informed consent before blood sampling Consent for publication Not applicable.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests:
The authors declare that they have no con ict of interest      Table 6 Correlation between content of Fe, Zn, Cr, Mn, Ca, Se, and trace elements in blood serum in the different groups