Studies illustrating the potential of artificial intelligence in multiomics
| Study | Genomics |
|---|---|
| Chen et al. [23] | A DL model, DeepVariant-AF, created by Google Health is applied to large data sets and reliably identifies gene variants. |
| Aradhya et al. [24] | DL models were trained using Invitae’s Evidence Modeling Platform to predict protein structure and function based on sequence data. |
| Kuenzi et al. [26] | A DL model, DrugCell, predicts the in vitro response of tumor cell lines to various drugs. |
| Sammut et al. [27] | A predictive model using combined sequence and digital pathology data predicts response to neo-adjuvant therapy in breast cancer patients. |
| Pathomics | |
| Källén et al. [29] | The DL model, OverFeat, accurately predicts the Gleason score using region-level tissue classification. |
| Hoang et al. [30] | A two-step model, ENLIGHT-DeepPT, predicts genome-wide tumor mRNA expression and treatment response to targeted and immune therapies based on digital pathology images of hematoxylin and eosin-stained (H&E) tumor slides. |
| Hu et al. [31] | CNN model predicts response to immune-checkpoint blockade based on the analysis of H&E slides alone. |
| Zhang et al. [32] | The DL model, PathoSig, predicts response to chemotherapy by identifying phenotypic clusters from H&E digital images. |
| Cheng et al. [33] | A DL model and three AI models quantitatively score PD-L1 expression. |
| Choi et al. [34] | A DL model improves the consensus of reads between pathologists and predicts response to treatment in patients with NSCLC. |
| Ligero et al. [35] | A Retrieval with Clustering-guided Contrastive Learning (RetCCL) model quantifies the degree of positivity of PD-L1 on IHC slides and predicts response to ICIs by estimating progression-free survival. |
| Radiomics | |
| Mu et al. [42] | A small residual convolutional network was employed to analyze PET/CT images and clinical data from NSCLC patients in order to develop a DL score that predicted PD-L1 expression. |
| He et al. [43] | Developed a novel non-invasive biomarker by integrating DL technology with CT characteristics to differentiate between NSCLC patients with high-TMB and low-TMB tumors and predict treatment efficacy. |
| Mu et al. [47] | Radiomic features from baseline pre-treatment 18F-FDG-PET/CT scans can predict clinical outcomes for NSCLC patients undergoing checkpoint blockade immunotherapy. |
| Vaidya et al. [48] | Radiomic markers extracted from baseline CT scans of advanced NSCLC patients treated with PD-1/PD-L1 inhibitors identifies patients at risk of hyperprogression. |
| Li et al. [49] | A CT-based radiomics model accurately predicts hyperprogression and pseudoprogression in NSCLC patients undergoing immunotherapy. |
| Metabolomics | |
| Curry et al. [51] | Developed an artificial neural network to help classify MS spectrometry. |
| Ball et al. [52] | Machine learning approaches to analyze MS data and identify metabolic signatures can differentiate patients with low versus high grade astrocytoma. |
| O’Shea et al. [54] | Use of an artificial neural network model with sputum metabolomics identifies six metabolites that were elevated in patients with small cell lung cancer compared to NSCLC. |
| Xie et al. [55] | Machine learning techniques identify six metabolites that distinguish stage 1 lung cancer patients from healthy controls. |
| Lipidomics | |
| Jiang et al. [63] | Lipidomic profiling identifies six key lipids, used to develop a predictive model for treatment response to chemo-immunotherapy. |
| Yu et al. [64] | Nine distinct lipids were used to predict immune related adverse events in NSCLC patients undergoing treatment with ICIs. |
| Immunogenomics | |
| Chen et al. [67] | Ground-glass associated lung cancers were less metabolically active and had a less active immune microenvironment compared to patients with solid lung nodules. |
| Sun et al. [68] | Immunohistochemistry and RNA-sequencing data show that NSCLC patients who lack either PD-L1 expression or immune infiltration may not benefit from immunotherapy. |
| Liu et al. [71] | Combined data from scanned histology slides and RNA-sequencing to develop an AI-based immunoscore model capable of predicting survival outcomes in patients with NSCLC who had received chemoimmunotherapy. |
| Breathomics | |
| Gordon et al. [76] | Gas chromatography-mass spectrometry (GC-MS) analysis of breath samples classifies 93% of patients with vs. without lung cancer. |
| Philipps et al. [77] | GC-MS analysis of 108 individuals confirms lung cancer in 60 patients. |
| Di Natale et al. [80] | Use of the electronic nose combined with partial least squares discriminant analysis correctly classifies 100% of lung cancer patients. |
| Mazzone et al. [82] | Colorimetric sensors accurately predict individuals with lung cancer versus individuals with other lung diseases. |
| Electronic health record | |
| Marmarelis et al. [84] | A nudge-based intervention with an EMR increases molecular testing and better guideline-concordant care. |
| Yuan et al. [85] | Developed a machine-learning-based prognostic model for NSCLC by the extraction of unstructured and structured data. |
DL: deep learning; CNN: convolutional neural network; NSCLC: non-small cell lung cancer; EMR: electronic medical record
