pathway informed multitask modeling

Current pathway-guided models lack a unified benchmark for simultaneously predicting eligibility for targeted therapy, need for radiotherapy, and six-month survival. This study proposes the first integrative evaluation framework based on Reactome pathway activity scores, jointly training three bioinformatic architectures—BINN, GraphPath, and PATH—across five TCGA cancer cohorts to enable multitask clinical outcome prediction. It innovatively applies deep learning over pathway structures to jointly model therapeutic response and survival, while establishing a cross-model protocol for fair comparison. Results show that PATH achieves overall superior performance in targeted therapy prediction, BINN excels in survival prediction, and GraphPath attains an AUROC of 0.92 for targeted therapy prediction in prostate cancer with well-defined driver mutations. Radiotherapy prediction remains suboptimal, likely because key decision-making factors are not captured in gene expression data.

This study addresses the challenge in traditional multi-task learning where heterogeneous output types—such as continuous and binary responses—lead to incomparable loss functions, hindering effective information sharing. To overcome this, the authors propose a multi-task transformation framework that unifies diverse response variables through unknown monotonic transformations. The approach integrates a shared first-layer deep neural network with group Lasso regularization, enabling joint modeling in high-dimensional settings. Notably, it is the first to combine monotonic transformations with a shared sparse structure, establishing a unified multi-task learning framework suitable for mixed output types. Theoretical guarantees for consistent variable selection are provided. Empirical results demonstrate that the method significantly outperforms existing approaches on both simulated data and real-world gene expression analysis, successfully identifying biologically meaningful shared predictors.

Estimating individual treatment effects and treatment–treatment interactions in multi-treatment settings faces two key challenges: insufficient parameter sharing across correlated treatments and exacerbated selection bias due to redundant latent variable modeling. To address these, we propose a unified framework integrating task embedding and balanced representation learning. A task embedding network enables parameter sharing across treatment modalities, while a nonparametric representation learning network—regularized by a learnable balancing penalty—avoids unnecessary latent variables, jointly mitigating confounding bias and selection bias. Our method synergistically combines deep learning, variational autoencoders, and learnable balancing constraints. In extensive synthetic experiments, it significantly outperforms state-of-the-art baselines. On real-world marketing data, it demonstrates high accuracy in estimating causal effects of multi-treatment combinations and strong practical deployability.

This study addresses the challenge of negative transfer in multitask learning with multimodal clinical data, which hinders effective modeling of related yet heterogeneous clinical outcomes. To overcome this limitation, the authors propose a unified Transformer-based multitask framework incorporating an Orthogonal Task Decomposition (OrthTD) mechanism. This approach explicitly decouples shared and task-specific subspaces at the representation level and enforces geometric orthogonality constraints to suppress redundancy and isolate task-unique signals. Evaluated on data from 12,430 surgical patients for predicting four distinct clinical outcomes, the model achieves an average AUC of 87.5% and AUPRC of 37.2%, significantly outperforming existing methods—particularly excelling in the detection of rare events.

Multimodal multitask prediction in clinical settings faces challenges from sample-level data heterogeneity—such as co-occurring structured rating scales and unstructured clinical text—and varying task interdependencies, compounded by pervasive missing values. Method: We propose the first sample-adaptive routing framework for unified multimodal multitask learning. Built upon a mixture-of-experts architecture, it jointly learns modality-specific processing paths (for raw/fused textual and numerical features) and task-sharing strategies (dynamically assigning shared or task-specific prediction heads), enabling personalized information flow. The model is trained end-to-end to yield interpretable awareness of modality importance and task relationships. Results: Evaluated on synthetic data and real-world psychotherapy transcripts, our method significantly outperforms fixed multitask and single-task baselines in predicting depression and anxiety symptom severity. It demonstrates improved personalization and cost-effectiveness for clinical interventions, validating its practical utility in mental health applications.

This study addresses the challenge of balancing model expressiveness and interpretability in multi-omics data integration by proposing a Pathway Activity Autoencoder (PAA). The PAA embeds prior biological pathway knowledge directly into the network architecture as structural constraints, thereby achieving intrinsic interpretability without sacrificing predictive power. By integrating diverse omics data—including gene expression, protein abundance, and miRNA profiles—and leveraging pathway-guided architectural design together with tailored regularization strategies, the method significantly outperforms existing approaches in breast cancer survival prediction and molecular subtype classification. Experimental results demonstrate not only enhanced predictive performance but also clear attribution of each omics layer’s contribution to the predictions, offering robustness and clinically meaningful interpretability.

This study addresses the significant variability in breast cancer patients’ response to neoadjuvant chemotherapy (NACT) and the urgent need for accurate prediction of pathological complete response (pCR). To this end, the authors propose a novel approach that, for the first time, employs a 3D spatiotemporal graph neural network to model longitudinal dynamic contrast-enhanced MRI (DCE-MRI) data. Their method explicitly captures temporal interactions across multiple imaging timepoints and incorporates three complementary self-supervised learning objectives to enable personalized treatment response prediction. Evaluated on the ISPY-2 dataset comprising 585 patients, the proposed framework substantially outperforms existing visual and self-supervised baselines, establishing a new state-of-the-art benchmark for pCR prediction. The authors further release their code and data processing library to support reproducible research in this domain.

This work addresses the frequent failure of multi-task Bayesian optimization due to inaccurate estimation of cross-task correlations—even under the simplified assumption of affine relationships between source and target tasks. The study systematically identifies two root causes: alignment errors induced by task normalization with finite samples and insufficient identifiability of marginal likelihood under non-overlapping experimental designs. To mitigate these issues, the authors propose modeling task-specific means and scales as learnable parameters and introduce three conservative remedies: enforcing non-negative task covariance constraints, adopting partially co-located experimental designs, and refining the correlation estimation mechanism. The method successfully recovers single-task performance on affine synthetic benchmarks and hyperparameter transfer tuning tasks, yet limitations persist in more complex settings, particularly those involving ranking-based objectives or latent contextual structures.