Automatic Extraction of Comprehensive Drug Safety Information from Adverse Drug Event Narratives in the Korea Adverse Event Reporting System Using Natural Language Processing Techniques.

INTRODUCTION: Concerns have been raised over the quality of drug safety information, particularly data completeness, collected through spontaneous reporting systems (SRS), although regulatory agencies routinely use SRS data to guide their pharmacovigilance programs. We expected that collecting additional drug safety information from adverse event (ADE) narratives and incorporating it into the SRS database would improve data completeness. OBJECTIVE: The aims of this study were to define the extraction of comprehensive drug safety information from ADE narratives reported through the Korea Adverse Event Reporting System (KAERS) as natural language processing (NLP) tasks and to provide baseline models for the defined tasks. METHODS: This study used ADE narratives and structured drug safety information from individual case safety reports (ICSRs) reported through KAERS between 1 January 2015 and 31 December 2019. We developed the annotation guideline for the extraction of comprehensive drug safety information from ADE narratives based on the International Conference on Harmonisation (ICH) E2B(R3) guideline and manually annotated 3723 ADE narratives. Then, we developed a domain-specific Korean Bidirectional Encoder Representations from Transformers (KAERS-BERT) model using 1.2 million ADE narratives in KAERS and provided baseline models for the task we defined. In addition, we performed an ablation experiment to investigate whether named entity recognition (NER) models were improved when a training dataset contained more diverse ADE narratives. RESULTS: We defined 21 types of word entities, six types of entity labels, and 49 types of relations to formulate the extraction of comprehensive drug safety information as NLP tasks. We obtained a total of 86,750 entities, 81,828 entity labels, and 45,107 relations from manually annotated ADE narratives. The KAERS-BERT model achieved F1-scores of 83.81 and 76.62% on the NER and sentence extraction tasks, respectively, while outperforming other baseline models on all the NLP tasks we defined except the sentence extraction task. Finally, utilizing the NER model for extracting drug safety information from ADE narratives resulted in an average increase of 3.24% in data completeness for KAERS structured data fields. CONCLUSIONS: We formulated the extraction of comprehensive drug safety information from ADE narratives as NLP tasks and developed the annotated corpus and strong baseline models for the tasks. The annotated corpus and models for extracting comprehensive drug safety information can improve the data quality of an SRS database.

Action-driven contrastive representation for reinforcement learning.

In reinforcement learning, reward-driven feature learning directly from high-dimensional images faces two challenges: sample-efficiency for solving control tasks and generalization to unseen observations. In prior works, these issues have been addressed through learning representation from pixel inputs. However, their representation faced the limitations of being vulnerable to the high diversity inherent in environments or not taking the characteristics for solving control tasks. To attenuate these phenomena, we propose the novel contrastive representation method, Action-Driven Auxiliary Task (ADAT), which forces a representation to concentrate on essential features for deciding actions and ignore control-irrelevant details. In the augmented state-action dictionary of ADAT, the agent learns representation to maximize agreement between observations sharing the same actions. The proposed method significantly outperforms model-free and model-based algorithms in the Atari and OpenAI ProcGen, widely used benchmarks for sample-efficiency and generalization.

Interpretable disease prediction using heterogeneous patient records with self-attentive fusion encoder.

OBJECTIVE: We propose an interpretable disease prediction model that efficiently fuses multiple types of patient records using a self-attentive fusion encoder. We assessed the model performance in predicting cardiovascular disease events, given the records of a general patient population. MATERIALS AND METHODS: We extracted 798111 ses and 67 623 controls from the sample cohort database and nationwide healthcare claims data of South Korea. Among the information provided, our model used the sequential records of medical codes and patient characteristics, such as demographic profiles and the most recent health examination results. These two types of patient records were combined in our self-attentive fusion module, whereas previously dominant methods aggregated them using a simple concatenation. The prediction performance was compared to state-of-the-art recurrent neural network-based approaches and other widely used machine learning approaches. RESULTS: Our model outperformed all the other compared methods in predicting cardiovascular disease events. It achieved an area under the curve of 0.839, while the other compared methods achieved between 0.74111 d 0.830. Moreover, our model consistently outperformed the other methods in a more challenging setting in which we tested the model's ability to draw an inference from more nonobvious, diverse factors. DISCUSSION: We also interpreted the attention weights provided by our model as the relative importance of each time step in the sequence. We showed that our model reveals the informative parts of the patients' history by measuring the attention weights. CONCLUSION: We suggest an interpretable disease prediction model that efficiently fuses heterogeneous patient records and demonstrates superior disease prediction performance.

Efficient spread-size approximation of opinion spreading in general social networks.

In contemporary society, understanding how information, such as trends and viruses, spreads in various social networks is an important topic in many areas. However, it is difficult to mathematically measure how widespread the information is, especially for a general network structure. There have been studies on opinion spreading, but many studies are limited to specific spreading models such as the susceptible-infected-recovered model and the independent cascade model, and it is difficult to apply these studies to various situations. In this paper, we first suggest a general opinion spreading model (GOSM) that generalizes a large class of popular spreading models. In this model, each node has one of several states, and the state changes through interaction with neighboring nodes at discrete time intervals. Next, we show that many GOSMs have a stable property that is a GOSM version of a uniform equicontinuity. Then, we provide an approximation method to approximate the expected spread size for stable GOSMs. For the approximation method, we propose a concentration theorem that guarantees that a generalized mean-field theorem calculates the expected spreading size within small error bounds for finite time steps for a slightly dense network structure. Furthermore, we prove that a "single simulation" of running the Monte Carlo simulation is sufficient to approximate the expected spreading size. We conduct experiments on both synthetic and real-world networks and show that our generalized approximation method well predicts the state density of the various models, especially in graphs with a large number of nodes. Experimental results show that the generalized mean-field approximation and a single Monte Carlo simulation converge as shown in the concentration theorem.

Hierarchical ordering with partial pairwise hierarchical relationships on the macaque brain data sets.

Hierarchical organizations of information processing in the brain networks have been known to exist and widely studied. To find proper hierarchical structures in the macaque brain, the traditional methods need the entire pairwise hierarchical relationships between cortical areas. In this paper, we present a new method that discovers hierarchical structures of macaque brain networks by using partial information of pairwise hierarchical relationships. Our method uses a graph-based manifold learning to exploit inherent relationship, and computes pseudo distances of hierarchical levels for every pair of cortical areas. Then, we compute hierarchy levels of all cortical areas by minimizing the sum of squared hierarchical distance errors with the hierarchical information of few cortical areas. We evaluate our method on the macaque brain data sets whose true hierarchical levels are known as the FV91 model. The experimental results show that hierarchy levels computed by our method are similar to the FV91 model, and its errors are much smaller than the errors of hierarchical clustering approaches.

Rumor Detection over Varying Time Windows.

This study determines the major difference between rumors and non-rumors and explores rumor classification performance levels over varying time windows-from the first three days to nearly two months. A comprehensive set of user, structural, linguistic, and temporal features was examined and their relative strength was compared from near-complete date of Twitter. Our contribution is at providing deep insight into the cumulative spreading patterns of rumors over time as well as at tracking the precise changes in predictive powers across rumor features. Statistical analysis finds that structural and temporal features distinguish rumors from non-rumors over a long-term window, yet they are not available during the initial propagation phase. In contrast, user and linguistic features are readily available and act as a good indicator during the initial propagation phase. Based on these findings, we suggest a new rumor classification algorithm that achieves competitive accuracy over both short and long time windows. These findings provide new insights for explaining rumor mechanism theories and for identifying features of early rumor detection.

Efficient Energy Minimization for Enforcing Label Statistics.

Energy minimization algorithms, such as graph cuts, enable the computation of the MAP solution under certain probabilistic models such as Markov random fields. However, for many computer vision problems, the MAP solution under the model is not the ground truth solution. In many problem scenarios, the system has access to certain statistics of the ground truth. For instance, in image segmentation, the area and boundary length of the object may be known. In these cases, we want to estimate the most probable solution that is consistent with such statistics, i.e., satisfies certain equality or inequality constraints. The above constrained energy minimization problem is NP-hard in general, and is usually solved using Linear Programming formulations, which relax the integrality constraints. This paper proposes a novel method that directly finds the discrete approximate solution of such problems by maximizing the corresponding Lagrangian dual. This method can be applied to any constrained energy minimization problem whose unconstrained version is polynomial time solvable, and can handle multiple, equality or inequality, and linear or non-linear constraints. One important advantage of our method is the ability to handle second order constraints with both-side inequalities with a weak restriction, not trivial in the relaxation based methods, and show that the restriction does not affect the accuracy in our cases.We demonstrate the efficacy of our method on the foreground/background image segmentation problem, and show that it produces impressive segmentation results with less error, and runs more than 20 times faster than the state-of-the-art LP relaxation based approaches.