skip to main content

Title: Can x2vec save lives? Integrating graph and language embeddings for automatic mental health classification

Graph and language embedding models are becoming commonplace in large scale analyses given their ability to represent complex sparse data densely in low-dimensional space. Integrating these models’ complementary relational and communicative data may be especially helpful if predicting rare events or classifying members of hidden populations—tasks requiring huge and sparse datasets for generalizable analyses. For example, due to social stigma and comorbidities, mental health support groups often form in amorphous online groups. Predicting suicidality among individuals in these settings using standard network analyses is prohibitive due to resource limits (e.g., memory), and adding auxiliary data like text to such models exacerbates complexity- and sparsity-related issues. Here, I show how merging graph and language embedding models (metapath2vecanddoc2vec) avoids these limits and extracts unsupervised clustering data without domain expertise or feature engineering. Graph and language distances to a suicide support group have little correlation (ρ< 0.23), implying the two models are not embedding redundant information. When used separately to predict suicidality among individuals, graph and language data generate relatively accurate results (69% and 76%, respectively) but have moderately large false-positive (25% and 21%, respectively) and false-negative (38% and 27%, respectively) rates; however, when integrated, both data produce highly accurate predictions (90%, more » with 10% false-positives and 12% false-negatives). Visualizing graph embeddings annotated with predictions of potentially suicidal individuals shows the integrated model could classify such individuals even if they are positioned far from the support group. These results extend research on the importance of simultaneously analyzing behavior and language in massive networks and efforts to integrate embedding models for different kinds of data when predicting and classifying, particularly when they involve rare events.

« less
Publication Date:
Journal Name:
Journal of Physics: Complexity
Page Range or eLocation-ID:
Article No. 035005
IOP Publishing
Sponsoring Org:
National Science Foundation
More Like this
  1. Public genomic repositories are notoriously lacking in racially and ethnically diverse samples. This limits the reaches of exploration and has in fact been one of the driving factors for the initiation of the All of Us project. Our particular focus here is to provide a model-based framework for accurately predicting DNA methylation from genetic data using racially sparse public repository data. Epigenetic alterations are of great interest in cancer research but public repository data is limited in the information it provides. However, genetic data is more plentiful. Our phenotype of interest is cervical cancer in The Cancer Genome Atlas (TCGA) repository. Being able to generate such predictions would nicely complement other work that has generated gene-level predictions of gene expression for normal samples. We develop a new prediction approach which uses shared random effects from a nested error mixed effects regression model. The sharing of random effects allows borrowing of strength across racial groups greatly improving predictive accuracy. Additionally, we show how to further borrow strength by combining data from different cancers in TCGA even though the focus of our predictions is DNA methylation in cervical cancer. We compare our methodology against other popular approaches including the elastic net shrinkagemore »estimator and random forest prediction. Results are very encouraging with the shared classified random effects approach uniformly producing more accurate predictions – overall and for each racial group.« less
  2. Abstract Motivation

    Computational methods for compound–protein affinity and contact (CPAC) prediction aim at facilitating rational drug discovery by simultaneous prediction of the strength and the pattern of compound–protein interactions. Although the desired outputs are highly structure-dependent, the lack of protein structures often makes structure-free methods rely on protein sequence inputs alone. The scarcity of compound–protein pairs with affinity and contact labels further limits the accuracy and the generalizability of CPAC models.


    To overcome the aforementioned challenges of structure naivety and labeled-data scarcity, we introduce cross-modality and self-supervised learning, respectively, for structure-aware and task-relevant protein embedding. Specifically, protein data are available in both modalities of 1D amino-acid sequences and predicted 2D contact maps that are separately embedded with recurrent and graph neural networks, respectively, as well as jointly embedded with two cross-modality schemes. Furthermore, both protein modalities are pre-trained under various self-supervised learning strategies, by leveraging massive amount of unlabeled protein data. Our results indicate that individual protein modalities differ in their strengths of predicting affinities or contacts. Proper cross-modality protein embedding combined with self-supervised learning improves model generalizability when predicting both affinities and contacts for unseen proteins.

    Availability and implementation

    Data and source codes are available at

    Supplementary information

    Supplementary data aremore »available at Bioinformatics online.

    « less
  3. Abstract

    Monitoring wildlife abundance across space and time is an essential task to study their population dynamics and inform effective management. Acoustic recording units are a promising technology for efficiently monitoring bird populations and communities. While current acoustic data models provide information on the presence/absence of individual species, new approaches are needed to monitor population abundance, ideally across large spatio‐temporal regions.

    We present an integrated modelling framework that combines high‐quality but temporally sparse bird point count survey data with acoustic recordings. Our models account for imperfect detection in both data types and false positive errors in the acoustic data. Using simulations, we compare the accuracy and precision of abundance estimates using differing amounts of acoustic vocalizations obtained from a clustering algorithm, point count data, and a subset of manually validated acoustic vocalizations. We also use our modelling framework in a case study to estimate abundance of the Eastern Wood‐Pewee (Contopus virens) in Vermont, USA.

    The simulation study reveals that combining acoustic and point count data via an integrated model improves accuracy and precision of abundance estimates compared with models informed by either acoustic or point count data alone. Improved estimates are obtained across a wide range of scenarios, with the largest gainsmore »occurring when detection probability for the point count data is low. Combining acoustic data with only a small number of point count surveys yields estimates of abundance without the need for validating any of the identified vocalizations from the acoustic data. Within our case study, the integrated models provided moderate support for a decline of the Eastern Wood‐Pewee in this region.

    Our integrated modelling approach combines dense acoustic data with few point count surveys to deliver reliable estimates of species abundance without the need for manual identification of acoustic vocalizations or a prohibitively expensive large number of repeated point count surveys. Our proposed approach offers an efficient monitoring alternative for large spatio‐temporal regions when point count data are difficult to obtain or when monitoring is focused on rare species with low detection probability.

    « less
  4. Transformer model architectures have revolutionized the natural language processing (NLP) domain and continue to produce state-of-the-art results in text-based applications. Prior to the emergence of transformers, traditional NLP models such as recurrent and convolutional neural networks demonstrated promising utility for patient-level predictions and health forecasting from longitudinal datasets. However, to our knowledge only few studies have explored transformers for predicting clinical outcomes from electronic health record (EHR) data, and in our estimation, none have adequately derived a health-specific tokenization scheme to fully capture the heterogeneity of EHR systems. In this study, we propose a dynamic method for tokenizing both discrete and continuous patient data, and present a transformer-based classifier utilizing a joint embedding space for integrating disparate temporal patient measurements. We demonstrate the feasibility of our clinical AI framework through multi-task ICU patient acuity estimation, where we simultaneously predict six mortality and readmission outcomes. Our longitudinal EHR tokenization and transformer modeling approaches resulted in more accurate predictions compared with baseline machine learning models, which suggest opportunities for future multimodal data integrations and algorithmic support tools using clinical transformer networks.
  5. Abstract Background

    Natural language processing (NLP) tasks in the health domain often deal with limited amount of labeled data due to high annotation costs and naturally rare observations. To compensate for the lack of training data, health NLP researchers often have to leverage knowledge and resources external to a task at hand. Recently, pretrained large-scale language models such as the Bidirectional Encoder Representations from Transformers (BERT) have been proven to be a powerful way of learning rich linguistic knowledge from massive unlabeled text and transferring that knowledge to downstream tasks. However, previous downstream tasks often used training data at such a large scale that is unlikely to obtain in the health domain. In this work, we aim to study whether BERT can still benefit downstream tasks when training data are relatively small in the context of health NLP.


    We conducted a learning curve analysis to study the behavior of BERT and baseline models as training data size increases. We observed the classification performance of these models on two disease diagnosis data sets, where some diseases are naturally rare and have very limited observations (fewer than 2 out of 10,000). The baselines included commonly used text classification models such as sparse andmore »dense bag-of-words models, long short-term memory networks, and their variants that leveraged external knowledge. To obtain learning curves, we incremented the amount of training examples per disease from small to large, and measured the classification performance in macro-averaged$$F_{1}$$F1score.


    On the task of classifying all diseases, the learning curves of BERT were consistently above all baselines, significantly outperforming them across the spectrum of training data sizes. But under extreme situations where only one or two training documents per disease were available, BERT was outperformed by linear classifiers with carefully engineered bag-of-words features.


    As long as the amount of training documents is not extremely few, fine-tuning a pretrained BERT model is a highly effective approach to health NLP tasks like disease classification. However, in extreme cases where each class has only one or two training documents and no more will be available, simple linear models using bag-of-words features shall be considered.

    « less