Uncovering expression signatures of synergistic drug responses via ensembles of explainable machine-learning models

A deep profile of gene expression across 18 human cancers

Clinically and biologically valuable information may reside untapped in large cancer gene expression data sets. Deep unsupervised learning has the potential to extract this information with unprecedented efficacy but has thus far been hampered by a lack of biological interpretability and robustness. Here, we present DeepProfile, a comprehensive framework that addresses current challenges in applying unsupervised deep learning to gene expression profiles. We use DeepProfile to learn low-dimensional latent spaces for 18 human cancers from 50,211 transcriptomes. DeepProfile outperforms existing dimensionality reduction methods with respect to biological interpretability. Using DeepProfile interpretability methods, we show that genes that are universally important in defining the latent spaces across all cancer types control immune cell activation, while cancer type-specific genes and pathways define molecular disease subtypes. By linking DeepProfile latent variables to secondary tumor characteristics, we discover that tumor mutation burden is closely associated with the expression of cell cycle-related genes. DNA mismatch repair and MHC class II antigen presentation pathway expression, on the other hand, are consistently associated with patient survival. We validate these results through Kaplan-Meier analyses and nominate tumor-associated macrophages as an important source of survival-correlated MHC class II transcripts. Our results illustrate the power of unsupervised deep learning for discovery of cancer biology from existing gene expression data.

Learning chemical sensitivity reveals mechanisms of cellular response

Chemical probes interrogate disease mechanisms at the molecular level by linking genetic changes to observable traits. However, comprehensive chemical screens in diverse biological models are impractical. To address this challenge, we develop ChemProbe, a model that predicts cellular sensitivity to hundreds of molecular probes and drugs by learning to combine transcriptomes and chemical structures. Using ChemProbe, we infer the chemical sensitivity of cancer cell lines and tumor samples and analyze how the model makes predictions. We retrospectively evaluate drug response predictions for precision breast cancer treatment and prospectively validate chemical sensitivity predictions in new cellular models, including a genetically modified cell line. Our model interpretation analysis identifies transcriptome features reflecting compound targets and protein network modules, identifying genes that drive ferroptosis. ChemProbe is an interpretable in silico screening tool that allows researchers to measure cellular response to diverse compounds, facilitating research into molecular mechanisms of chemical sensitivity.

DrugMAP 2.0: molecular atlas and pharma-information of all drugs

The escalating costs and high failure rates have decelerated the pace of drug development, which amplifies the research interests in developing combinatorial/repurposed drugs and understanding off-target adverse drug reaction (ADR). In other words, it is demanded to delineate the molecular atlas and pharma-information for the combinatorial/repurposed drugs and off-target interactions. However, such invaluable data were inadequately covered by existing databases. In this study, a major update was thus conducted to the DrugMAP, which accumulated (a) 20831 combinatorial drugs and their interacting atlas involving 1583 pharmacologically important molecules; (b) 842 repurposed drugs and their interacting atlas with 795 molecules; (c) 3260 off-targets relevant to the ADRs of 2731 drugs and (d) various types of pharmaceutical information, including diverse ADMET properties, versatile diseases, and various ADRs/off-targets. With the growing demands for discovering combinatorial/repurposed therapies and the rapidly emerging interest in AI-based drug discovery, DrugMAP was highly expected to act as an indispensable supplement to existing databases facilitating drug discovery, which was accessible at: https://idrblab.org/drugmap/.

An end-to-end deep learning method for mass spectrometry data analysis to reveal disease-specific metabolic profiles

Untargeted metabolomic analysis using mass spectrometry provides comprehensive metabolic profiling, but its medical application faces challenges of complex data processing, high inter-batch variability, and unidentified metabolites. Here, we present DeepMSProfiler, an explainable deep-learning-based method, enabling end-to-end analysis on raw metabolic signals with output of high accuracy and reliability. Using cross-hospital 859 human serum samples from lung adenocarcinoma, benign lung nodules, and healthy individuals, DeepMSProfiler successfully differentiates the metabolomic profiles of different groups (AUC 0.99) and detects early-stage lung adenocarcinoma (accuracy 0.961). Model flow and ablation experiments demonstrate that DeepMSProfiler overcomes inter-hospital variability and effects of unknown metabolites signals. Our ensemble strategy removes background-category phenomena in multi-classification deep-learning models, and the novel interpretability enables direct access to disease-related metabolite-protein networks. Further applying to lipid metabolomic data unveils correlations of important metabolites and proteins. Overall, DeepMSProfiler offers a straightforward and reliable method for disease diagnosis and mechanism discovery, enhancing its broad applicability.

Synergistic Machine Learning Accelerated Discovery of Nanoporous Inorganic Crystals as Non-Absorbable Oral Drugs

Machine learning (ML) has taken drug discovery to new heights, where effective ML training requires vast quantities of high-quality experimental data as input. Non-absorbable oral drugs (NODs) have unique safety advantage for chronic diseases due to their zero systemic exposure, but their empirical discovery is still time-consuming and costly. Here, a synergistic ML method, integrating small data-driven multi-layer unsupervised learning, in silico quantum-mechanical computations, and minimal wet-lab experiments is devised to identify the finest NODs from massive inorganic materials to achieve multi-objective function (high selectivity, large capacity, and stability). Based on this method, a NH4-form nanoporous zeolite with merlinoite (MER) framework (NH4-MER) is discovered for the treatment of hyperkalemia. In three different animal models, NH4-MER shows a superior safety and efficacy profile in reducing blood K+ without Na+ release, which is an unmet clinical need in chronic kidney disease and Gordon's syndrome. This work provides a synergistic ML method to accelerate the discovery of NODs and other shape-selective materials.

Digital assessment of tertiary lymphoid structures and therapeutic responses in gastric cancer: A multi-centric retrospective study

BACKGROUND

Tertiary lymphoid structures (TLSs) are associated with favorable prognosis and enhanced response to anti-cancer therapy. A digital assessment of TLSs could provide an objective alternative that mitigates variability inherent in manual evaluation. This study aimed to develop and validate a digital gene panel based on biological prior knowledge for assessment of TLSs, and further investigate its associations with survival and multiple anti-cancer therapies.

MATERIALS AND METHODS

The present study involved 1,704 patients with gastric cancer from seven cancer centers. TLSs were identified morphologically through hematoxylin-and-eosin staining. We further developed a digital score based on targeted gene expression profiling to assess TLSs status, recorded as gene signature of tertiary lymphoid structures (gsTLS). For enhanced interpretability, we employed the SHapley Additive exPlanation (SHAP) analysis to elucidate its contribution to the prediction. We next evaluated the signature's associations with prognosis, and investigated its predictive accuracy for multiple anti-cancer therapies, including adjuvant chemotherapy and immunotherapy.

RESULTS

The gsTLS panel with nine gene features achieved high accuracies in predicting TLSs status in the training, internal and external validation cohorts (area under the curve, range: 0.729-0.791). In multivariable analysis, gsTLS remained an independent predictor of disease-free and overall survival (hazard ratio, range: 0.346-0.743, all P < 0.05) after adjusting for other clinicopathological variables. SHAP analysis highlighted gsTLS as the strongest predictor of TLSs status compared with clinical features. Importantly, patients with high gsTLS (but not those with low gsTLS) exhibited substantial benefits from adjuvant chemotherapy (P < 0.05). Furthermore, we found that the objective response rate to anti-programmed cell death protein 1 (anti-PD-1) immunotherapy was significantly higher in the high-gsTLS group (40.7%) versus the low-gsTLS group (5.6%, P = 0.036), and the diagnosis was independent from Epstein-Barr virus (EBV), tumor mutation burden (TMB), and programmed cell death-ligand 1 (PD-L1) expression.

CONCLUSION

The gsTLS digital panel enables accurate assessment of TLSs status, and provides information regarding prognosis and responses to multiple therapies for gastric cancer.

A guide to artificial intelligence for cancer researchers

Artificial intelligence (AI) has been commoditized. It has evolved from a specialty resource to a readily accessible tool for cancer researchers. AI-based tools can boost research productivity in daily workflows, but can also extract hidden information from existing data, thereby enabling new scientific discoveries. Building a basic literacy in these tools is useful for every cancer researcher. Researchers with a traditional biological science focus can use AI-based tools through off-the-shelf software, whereas those who are more computationally inclined can develop their own AI-based software pipelines. In this article, we provide a practical guide for non-computational cancer researchers to understand how AI-based tools can benefit them. We convey general principles of AI for applications in image analysis, natural language processing and drug discovery. In addition, we give examples of how non-computational researchers can get started on the journey to productively use AI in their own work.

MFSynDCP: multi-source feature collaborative interactive learning for drug combination synergy prediction

Drug combination therapy is generally more effective than monotherapy in the field of cancer treatment. However, screening for effective synergistic combinations from a wide range of drug combinations is particularly important given the increase in the number of available drug classes and potential drug-drug interactions. Existing methods for predicting the synergistic effects of drug combinations primarily focus on extracting structural features of drug molecules and cell lines, but neglect the interaction mechanisms between cell lines and drug combinations. Consequently, there is a deficiency in comprehensive understanding of the synergistic effects of drug combinations. To address this issue, we propose a drug combination synergy prediction model based on multi-source feature interaction learning, named MFSynDCP, aiming to predict the synergistic effects of anti-tumor drug combinations. This model includes a graph aggregation module with an adaptive attention mechanism for learning drug interactions and a multi-source feature interaction learning controller for managing information transfer between different data sources, accommodating both drug and cell line features. Comparative studies with benchmark datasets demonstrate MFSynDCP's superiority over existing methods. Additionally, its adaptive attention mechanism graph aggregation module identifies drug chemical substructures crucial to the synergy mechanism. Overall, MFSynDCP is a robust tool for predicting synergistic drug combinations. The source code is available from GitHub at https://github.com/kkioplkg/MFSynDCP .

Network pharmacology: towards the artificial intelligence-based precision traditional Chinese medicine

Network pharmacology (NP) provides a new methodological perspective for understanding traditional medicine from a holistic perspective, giving rise to frontiers such as traditional Chinese medicine network pharmacology (TCM-NP). With the development of artificial intelligence (AI) technology, it is key for NP to develop network-based AI methods to reveal the treatment mechanism of complex diseases from massive omics data. In this review, focusing on the TCM-NP, we summarize involved AI methods into three categories: network relationship mining, network target positioning and network target navigating, and present the typical application of TCM-NP in uncovering biological basis and clinical value of Cold/Hot syndromes. Collectively, our review provides researchers with an innovative overview of the methodological progress of NP and its application in TCM from the AI perspective.

Explainable artificial intelligence for omics data: a systematic mapping study

Researchers increasingly turn to explainable artificial intelligence (XAI) to analyze omics data and gain insights into the underlying biological processes. Yet, given the interdisciplinary nature of the field, many findings have only been shared in their respective research community. An overview of XAI for omics data is needed to highlight promising approaches and help detect common issues. Toward this end, we conducted a systematic mapping study. To identify relevant literature, we queried Scopus, PubMed, Web of Science, BioRxiv, MedRxiv and arXiv. Based on keywording, we developed a coding scheme with 10 facets regarding the studies' AI methods, explainability methods and omics data. Our mapping study resulted in 405 included papers published between 2010 and 2023. The inspected papers analyze DNA-based (mostly genomic), transcriptomic, proteomic or metabolomic data by means of neural networks, tree-based methods, statistical methods and further AI methods. The preferred post-hoc explainability methods are feature relevance (n = 166) and visual explanation (n = 52), while papers using interpretable approaches often resort to the use of transparent models (n = 83) or architecture modifications (n = 72). With many research gaps still apparent for XAI for omics data, we deduced eight research directions and discuss their potential for the field. We also provide exemplary research questions for each direction. Many problems with the adoption of XAI for omics data in clinical practice are yet to be resolved. This systematic mapping study outlines extant research on the topic and provides research directions for researchers and practitioners.

Learning chemical sensitivity reveals mechanisms of cellular response

Chemical probes interrogate disease mechanisms at the molecular level by linking genetic changes to observable traits. However, comprehensive chemical screens in diverse biological models are impractical. To address this challenge, we developed ChemProbe, a model that predicts cellular sensitivity to hundreds of molecular probes and drugs by learning to combine transcriptomes and chemical structures. Using ChemProbe, we inferred the chemical sensitivity of cancer cell lines and tumor samples and analyzed how the model makes predictions. We retrospectively evaluated drug response predictions for precision breast cancer treatment and prospectively validated chemical sensitivity predictions in new cellular models, including a genetically modified cell line. Our model interpretation analysis identified transcriptome features reflecting compound targets and protein network modules, identifying genes that drive ferroptosis. ChemProbe is an interpretable in silico screening tool that allows researchers to measure cellular response to diverse compounds, facilitating research into molecular mechanisms of chemical sensitivity.

Algorithmic fairness in artificial intelligence for medicine and healthcare

In healthcare, the development and deployment of insufficiently fair systems of artificial intelligence (AI) can undermine the delivery of equitable care. Assessments of AI models stratified across subpopulations have revealed inequalities in how patients are diagnosed, treated and billed. In this Perspective, we outline fairness in machine learning through the lens of healthcare, and discuss how algorithmic biases (in data acquisition, genetic variation and intra-observer labelling variability, in particular) arise in clinical workflows and the resulting healthcare disparities. We also review emerging technology for mitigating biases via disentanglement, federated learning and model explainability, and their role in the development of AI-based software as a medical device.