DeePaN: deep patient graph convolutional network integrating clinico-genomic evidence to stratify lung cancers for immunotherapy

Abstract

Immuno-oncology (IO) therapies have transformed the therapeutic landscape of non-small cell lung cancer (NSCLC). However, patient responses to IO are variable and influenced by a heterogeneous combination of health, immune, and tumor factors. There is a pressing need to discover the distinct NSCLC subgroups that influence response. We have developed a deep patient graph convolutional network, we call “DeePaN”, to discover NSCLC complexity across data modalities impacting IO benefit. DeePaN employs high-dimensional data derived from both real-world evidence (RWE)-based electronic health records (EHRs) and genomics across 1937 IO-treated NSCLC patients. DeePaN demonstrated effectiveness to stratify patients into subgroups with significantly different (P-value of 2.2 × 10−11) overall median survival of 20.35 months and 9.42 months post-IO therapy. Significant differences in IO outcome were not seen from multiple non-graph-based unsupervised methods. Furthermore, we demonstrate that patient stratification from DeePaN has the potential to augment the emerging IO biomarker of tumor mutation burden (TMB). Characterization of the subgroups discovered by DeePaN indicates potential to inform IO therapeutic insight, including the enrichment of mutated KRAS and high blood monocyte count in the IO beneficial and IO non-beneficial subgroups, respectively. Our work has proven the concept that graph-based AI is feasible and can effectively integrate high-dimensional genomic and EHR data to meaningfully stratify cancer patients on distinct clinical outcomes, with potential to inform precision oncology.

DSTG: deconvoluting spatial transcriptomics data through graph-based artificial intelligence

“…Through learning the shared kernel used in spectral graph convolution across all nodes in a graph, a semi-supervised GCN model captures local graph structures as well as node features and incorporates both information as latent space representation. GCN [ 19 ] and its variants [ 20 , 21 ] have been applied to different scenarios successfully, including cancer patient subtyping using real-world evidence [ 22 ], protein prediction [ 23 ] and drug design [ 24 ], as well as single cells and diseases [ 25–29 ]. These works shows that, through effectively learning and leveraging the latent representation and topological relations among data, GCN models are able to significantly improve learning performance.…”

Artificial intelligence in early drug discovery enabling precision medicine

“…Therefore, they often neglect complementary information and dependencies contained in different molecular layers [91]. To address this, attempts have been directed to the development of machine learning techniques in multi-omics integration [23]. These methods jointly model data from various molecular sources [92] and extract directions of common variance using, e.g.…”

“…The identification of actionable disease subtypes is increasingly powered by AI using integrative methods combining diverse data modalities [17][18][19][20][21][22][23][24][25][26]. Complementarily, pharmacogenomics benefits from machine learning methods predicting in vitro monotherapy response by using molecular data in cell cultures, which may yield novel predictive biomarkers through the interpretation of the learned models [27][28][29][30][31][32][33][34].…”

An Unsupervised Graph Embeddings Approach to Multiplex Immunofluorescence Image Exploration

“…In [5] the DeePaN framework is described, where unsupervised learning is carried out on graphs integrating genomics and electronic health records data together, in order to identify responsive and non-responsive patient subsets amongst IO therapies addressing Non Small Cell Lung Cancer (NSCLC).…”