Explainable machine-learning algorithms to differentiate bipolar disorder from major depressive disorder using self-reported symptoms, vital signs, and blood-based markers.

BACKGROUND AND OBJECTIVE: Caused by shared genetic risk factors and similar neuropsychological symptoms, bipolar disorder (BD) and major depressive disorder (MDD) are at high risk of misdiagnosis, which is associated with ineffective treatment and worsening of outcomes. We aimed to develop a machine learning (ML)-based diagnostic system, based on electronic medical records (EMR) data, to mimic the clinical reasoning of human physicians to differentiate MDD and BD (especially BD depressive episodes) patients about to be admitted to a hospital and, hence, reduce the misdiagnosis of BD as MDD on admission. In addition, we examined to what extent our ML model could be made interpretable by quantifying and visualizing the features that drive the predictions. METHODS: By identifying 16,311 patients admitted to a hospital located in western China between 2009 and 2018 with a recorded main diagnosis of MDD or BD, we established three sub-cohorts with different combinations of features for both the MDD-BD cohort and the MDD-BD depressive episodes cohort, respectively. Four different ML algorithms (logistic regression, extreme gradient boosting (XGBoost), random forest, and support vector machine) and four train-test splits were used to train and validate diagnostic models, and explainable methods (SHAP and Break Down) were utilized to analyze the contribution of each of the features at both population-level and individual-level, including feature importance, feature interaction, and feature effect on prediction decision for a specific subject. RESULTS: The XGBoost algorithm provided the best test performance (AUC: 0.838 (0.810-0.867), PPV: 0.810 and NPV: 0.834) for separating patients with BD from those with MDD. Core predictors included symptoms (mood-up, exciting, bad sleep, loss of interest, talking, mood-down, provoke), along with age, job, myocardial enzyme markers (creatine kinase, hydroxybutyrate dehydrogenase), diabetes-associated marker (glucose), bone function marker (alkaline phosphatase), non-enzymatic antioxidant (uric acid), markers of immune/inflammation (white blood cell count, lymphocyte count, basophil percentage, monocyte count), cardiovascular function marker (low density lipoprotein), renal marker (total protein), liver biochemistry marker (indirect bilirubin), and vital signs like pulse. For separating patients with BD depressive episodes from those with MDD, the test AUC was 0.777 (0.732-0.822), with PPV 0.576 and NPV 0.899. Additional validation in models built with self-reported symptoms removed from the feature set, showed test AUC of 0.701 (0.666-0.736) for differentiating BD and MDD, and AUC of 0.564 (0.515-0.614) for detecting patients in BD depressive episodes from MDD patients. Validation in the datasets without removing the patients with comorbidity showed an AUC of 0.826 (0.806-0.846). CONCLUSION: The diagnostic system accurately identified patients with BD in various clinical scenarios, and differences in patterns of peripheral markers between BD and MDD could enrich our understanding of potential underlying pathophysiological mechanisms of them.

Dissecting clinical and biological heterogeneity in clinical states of bipolar disorder: a 10-year retrospective study from China.

Objectives: To dissect clinical and biological heterogeneity in clinical states of bipolar disorder (BD), and investigate if neuropsychological symptomatology, comorbidity, vital signs, and blood laboratory indicators are predictors of distinct BD states. Methods: A retrospective BD cohort was established with data extracted from a Chinese hospital's electronic medical records (EMR) between 2009 and 2018. Subjects were inpatients with a main discharge diagnosis of BD and were assessed for clinical state at hospitalization. We categorized all subjects into manic state, depressive state, and mixed state. Four machine learning classifiers were utilized to classify the subjects. A Shapley additive explanations (SHAP) algorithm was applied to the classifiers to aid in quantifying and visualizing the contributions of each feature that drive patient-specific classifications. Results: A sample of 3,085 records was included (38.54% as manic, 56.69% as depressive, and 4.77% as mixed state). Mixed state showed more severe suicidal ideation and psychomotor abnormalities, while depressive state showed more common anxiety, sleep, and somatic-related symptoms and more comorbid conditions. Higher levels of body temperature, pulse, and systolic and diastolic blood pressures were present during manic episodes. Xgboost achieved the best AUC of 88.54% in manic/depressive states classification; Logistic regression and Random forest achieved the best AUCs of 75.5 and 75% in manic/mixed states and depressive/mixed states classifications, respectively. Myocardial enzymes and the non-enzymatic antioxidant uric acid and bilirubin contributed significantly to distinguish BD clinical states. Conclusion: The observed novel biological associations with BD clinical states confirm that biological heterogeneity contributes to clinical heterogeneity of BD.

Cost-Effective Machine Learning Based Clinical Pre-Test Probability Strategy for DVT Diagnosis in Neurological Intensive Care Unit.

In order to overcome the shortage of the current costly DVT diagnosis and reduce the waste of valuable healthcare resources, we proposed a new diagnostic approach based on machine learning pre-test prediction models using EHRs. We examined the sociodemographic and clinical factors in the prediction of DVT with 518 NICU admitted patients, including 189 patients who eventually developed DVT. We used cross-validation on the training data to determine the optimal parameters, and finally, the applied ROC analysis is adopted to evaluate the predictive strength of each model. Two models (GLM and SVM) with the strongest ROC were selected for DVT prediction, based on which, we optimized the current intervention and diagnostic process of DVT and examined the performance of the proposed approach through simulations. The use of machine learning based pre-test prediction models can simplify and improve the intervention and diagnostic process of patients in NICU with suspected DVT, and reduce the valuable healthcare resource occupation/usage and medical costs.

Tuning the Light Response of Strong Polyelectrolyte Brushes with Counterions.

In this work, a negatively charged poly(3-sulfopropyl methacrylate potassium) (PSPMA) brush has been employed as a model system to demonstrate the tuning of the light response of strong polyelectrolyte brushes (SPBs) with counterions. The substitution of K+ counterions by azobenzene-containing counterions (Azo-N+) renders the PSPMA brush light-responsive in aqueous solutions. Nevertheless, the strength of the light response of the PSPMA brush is weak due to the inefficient disassembly of the micelle-like aggregates in the brush upon irradiation with ultraviolet light. Counterion mixtures of Azo-N+ and K+ are employed to realize a strong light response of the PSPMA brush by incorporating a reasonable amount of Azo-N+ counterions into the brush. The strength of the light response of the PSPMA brush can be tuned by the mole ratio of Azo-N+ to K+. Furthermore, properties including the hydration and conformation of the PSPMA brush can be reversibly switched via alternating ultraviolet and visible light irradiation. This work opens up the opportunities available for the use of counterions to tune the light response of SPBs.

Counterion-Tunable Thermosensitivity of Strong Polyelectrolyte Brushes.

In this work, poly(sodium styrene sulfonate) brushes have been employed as a precursor to prepare thermosensitive strong polyelectrolyte brushes (SPBs) through a counterion exchange strategy. The substitution of hydrophilic Na+ counterions by hydrophobic tetraalkylphosphonium counterions leads to a thermoresponsivity of the SPBs. The thermosensitive properties including hydration, stiffness, and surface water wettability of the SPBs can be modulated by the type of the tetraalkylphosphonium counterions. Nevertheless, the wet thickness of the SPBs with tetraalkylphosphonium counterions does not exhibit an obvious temperature dependency due to the high steric barrier in the crowded environment of SPBs generated by the large tetraalkylphosphonium counterions. The mixtures of small Na+ counterions and large tetraalkylphosphonium counterions are employed to realize the thermosensitive wet thickness without sacrificing other thermoresponsive properties of the SPBs because the mixed counterions can bring both a certain hydrophobicity and some free space to the brushes. This work opens up the opportunities available for the use of counterions to tune the thermosensitivity of SPBs.

Controlling Hydrogel Properties by the Cooperative Harvesting of the Components of Sunlight.

The utilization of the full spectrum of sunlight from ultraviolet to infrared to cooperatively control the properties of hydrogels is the key to developing efficient sunlight-responsive hydrogels, but there are significant challenges. The photoresponsive hydrogel developed here formed by azobenzene-containing polyzwitterionic chains possesses capabilities for simultaneously utilizing light from the ultraviolet to the infrared to cooperatively facilitate the gel-to-sol transitions. The ultraviolet and visible light can be converted into mechanical energy simultaneously to synergistically power the dynamic wagging of azobenzene groups, while the infrared light can be converted into kinetic energy of the zwitterionic groups to promote the wagging of the azobenzene groups. Cooperative action between the major components of sunlight can be applied to control a range of important properties of the hydrogel including wettability, adhesion, molecular release, self-healing, and mineralization. The work presented here demonstrates a novel strategy for harvesting all the major components of sunlight to control the properties of photoresponsive materials.

Counterion Specificity of Polyelectrolyte Brushes: Role of Specific Ion-Pairing Interactions.

We demonstrate here that the properties of poly (2-(methacryloyloxy) ethyl trimethylammonium chloride) brushes can be tuned by counterion species. When the brushes are exposed to external chloride (Cl- ) counterions, obvious dehydration and collapse are only observed at high salt concentrations. In the presence of very strongly chaotropic perchlorate (ClO4- ), the brushes strongly dehydrate and collapse at a very low salt concentration. For the strongly chaotropic thiocyanate ion (SCN- ), the changes in hydration and conformation of the brushes are similar to those observed for ClO4- but at a smaller extent at very low salt concentrations. With the addition of kosmotropic acetate (Ac- ), hydration of the brushes increases, accompanied by a swelling of the brushes in the low-salt-concentration regime. In contrast, the brushes dehydrate and collapse with increasing concentration of Ac- in the high-salt-concentration regime. The counterion specificity of the brushes demonstrated here is determined by specific ion-pairing interactions through modulating the osmotic pressure within the brushes and the hydrophobicity of the ion pairs.

Effect of Salt Concentration on the pH Responses of Strong and Weak Polyelectrolyte Brushes.

Strong polyelectrolyte brushes (SPB) and weak polyelectrolyte brushes (WPB) have different origins with response to pH, which makes their pH-responsive properties sensitive to salt concentration in different ways. Herein, we have employed poly[2-(methacryloyloxy)ethyl trimethylammonium chloride] (PMETAC) and poly[2-(dimethylamino)ethyl methacrylate] (PDMAEMA) brushes as model systems for SPB and WPB, respectively, to investigate the effect of salt concentration on the pH responses of SPB and WPB using several surface-sensitive techniques. The pH-responsive properties of the PMETAC brushes are governed by the reorganization of the interchain hydrogen bonds between the grafted chains, whereas the pH response of the PDMAEMA brushes is controlled by the charge of the grafted chains. The response of the properties of the PMETAC brushes including hydration, conformation, and surface wettability becomes weaker with increasing salt concentration induced by the competitive adsorption of counterions to the brushes between OH- and Cl-. The weakening of the pH response of the PMETAC brushes is more remarkable at the relatively high pH values. The pH response of the PDMAEMA brushes also exhibits a salt-concentration dependence. As the salt concentration increases, the weakening of the pH response of the PDMAEMA brushes is attributed to the decrease in osmotic pressure within the brushes at relatively low pH values.

Ion specificities of artificial macromolecules.

Artificial macromolecules are well-defined synthetic polymers, with a relatively simple structure as compared to naturally occurring macromolecules. This review focuses on the ion specificities of artifical macromolecules. Ion specificities are influenced by solvent-mediated indirect ion-macromolecule interactions and also by direct ion-macromolecule interactions. In aqueous solutions, the role of water-mediated indirect ion-macromolecule interactions will be discussed. The addition of organic solvents to aqueous solutions significantly changes the ion specificities due to the formation of water-organic solvent complexes. For direct ion-macromolecule interactions, we will discuss specific ion-pairing interactions for charged macromolecules and specific ion-neutral site interactions for uncharged macromolecules. When the medium conditions change from dilute solutions to crowded environments, the ion specificities can be modified by either the volume exclusion effect, the variation of dielectric constant, or the interactions between ions, macromolecules, and crowding agents.

Reorganization of hydrogen bond network makes strong polyelectrolyte brushes pH-responsive.

Weak polyelectrolytes have found extensive practical applications owing to their rich pH-responsive properties. In contrast, strong polyelectrolytes have long been regarded as pH-insensitive based on the well-established fact that the average degree of charging of strong polyelectrolyte chains is independent of pH. The possible applications of strong polyelectrolytes in smart materials have, thus, been severely limited. However, we demonstrate that almost all important properties of strong polyelectrolyte brushes (SPBs), such as chain conformation, hydration, stiffness, surface wettability, lubricity, adhesion, and protein adsorption are sensitive to pH. The pH response originates from the reorganization of the interchain hydrogen bond network between the grafted chains, triggered by the pH-mediated adsorption-desorption equilibrium of hydronium or hydroxide with the brushes. The reorganization process is firmly identified by advanced sum-frequency generation vibrational spectroscopy. Our findings not only provide a new understanding of the fundamental properties of SPBs but also uncover an extensive family of building blocks for constructing pH-responsive materials.

Anion Specificity of Polyzwitterionic Brushes with Different Carbon Spacer Lengths and Its Application for Controlling Protein Adsorption.

Both ion-specific interaction and carbon spacer length have strong effects on the properties of polyzwitterions. In this work, we have investigated the anion specificity of poly(sulfobetaine methacrylamide) (PSBMAm) brushes with different carbon spacer lengths. The effectiveness of anions to enhance the hydration of the PSBMAm brushes increases from kosmotropic to chaotropic anions. The interactions between the anions and the PSBMAm brushes are strongly influenced by carbon spacer length because the strength of inter/intrachain association of the PSBMAm brushes decreases with increasing carbon spacer length. The inter/intrachain association of the PSBMAm brushes with a longer carbon spacer is easier to break by the external anions in the high salt concentration regime. On the other hand, a longer carbon spacer is more favorable for the zwitterionic groups to form cyclic intramolecular structures. As a result, the addition of anions can more effectively enhance the hydration of the PSBMAm brushes with a medium-length carbon spacer compared with that of the PSBMAm brushes with a either shorter or longer carbon spacer in the low salt concentration regime, determined by the balance between the inter/intrachain association and the formation of cyclic intramolecular structures. Our study also demonstrates that both anion identity and carbon spacer length can be used to control protein adsorption on the surface of the PSBMAm brushes.

Interactions between Polyelectrolyte Brushes and Hofmeister Ions: Chaotropes versus Kosmotropes.

We have investigated the interactions between the positively charged poly[2-(methacryloyloxy)ethyltrimethylammonium chloride] (PMETAC) brushes and the Hofmeister anions and the interactions between the negatively charged poly(3-sulfopropyl methacrylate potassium) (PSPMA) brushes and the Hofmeister cations using a combination of quartz crystal microbalance with dissipation and spectroscopic ellipsometry. A V-shaped anion series is observed in terms of the ion-specific interactions between the PMETAC brushes and the Hofmeister anions. We have found that the chaotropic and kosmotropic anions interact with the PMETAC brushes in different manners. The ion-specific interactions between the PMETAC brushes and the chaotropic anions are dominated by the direct interactions between the anions and the positively charged quaternary ammonium group via ion pairing mediated by ionic hydration strength or polarizability, whereas the ion-specific interactions between the PMETAC brushes and the kosmotropic anions are dominated by the competition for water molecules between the anions and the brushes. The ion-specific interactions between the PMETAC brushes and the anions have significant influences on both the hydration and the conformation of the brushes. The cations exhibit weaker specific ion effects on the PSPMA brushes in comparison with the specific anion effects on the PMETAC brushes.