333 - A Digital Pathology Multimodal Artificial Intelligence Algorithm is Associated with Pro-Metastatic Genomic Pathways in Oligometastatic Prostate Cancer

Y. Song¹, A. C. Shetty², P. Sutera³, M. P. Deek⁴, A. Mendes³, K. Van der Eecken⁵, E. Chen⁶, T. N. Showalter⁷, T. J. Royce⁸, T. Todorovic⁹, H. C. Huang¹⁰, S. Houck⁶, R. Yamashita⁶, A. P. Kiess¹¹, D. Song³, T. L. Lotan¹², A. Esteva¹³, F. Y. Feng¹⁴, P. Ost¹⁵, and P. T. Tran¹⁶; ¹Department of Radiation Oncology, Division of Translational Radiation Sciences, University of Maryland Baltimore, School of Medicine,, Baltimore, MD, ²University of Maryland, Baltimore, MD, ³Johns Hopkins University School of Medicine, Baltimore, MD, ⁴Rutgers Cancer Institute of New Jersey, Department of Radiation Oncology, New Brunswick, NJ, ⁵Ghent University, Ghent, Belgium, ⁶Artera, Los Altos, CA, ⁷University of Virginia, Charlottesville, VA, ⁸The University of North Carolina at Chapel Hill, Chapel Hill, NC, ⁹Artera, Menlo Park, CA, United States, ¹⁰Decipher Biosciences (Vancouver, BC), Vancouver, Canada, ¹¹Department of Radiation Oncology and Molecular Radiation Sciences, Johns Hopkins University School of Medicine, Baltimore, MD, ¹²Department of Pathology, Johns Hopkins University School of Medicine, Baltimore, MD, ¹³Artera, Mountain View, CA, ¹⁴Department of Radiation Oncology, University of California San Francisco, San Francisco, CA, ¹⁵Department of Radiation Oncology, Ghent University Hospital, Ghent, Belgium, ¹⁶Department of Radiation Oncology, University of Maryland School of Medicine, Baltimore, MD

Purpose/Objective(s): Oligometastatic castration-sensitive prostate cancer (omCSPC) is a heterogeneous disease. Recently, a multimodal artificial intelligence (MMAI) algorithm (ArteraAI Prostate Test), which incorporates digital pathology and clinical information, has been shown to be prognostic in localized prostate cancer. Here, we evaluated the association between the MMAI score and vector features (VF) of the score with genomics in omCSPC. Materials/

Methods: We correlated somatic pathogenic mutations and ArteraAI MMAI scores from 168 omCSPC patients (133 metachronous and 35 synchronous). RNAseq profiling was performed on a subset of 65 metachronous patients. Somatic nonsynonymous pathogenic mutations from panel DNAseq were identified using Mutect2 and ClinVAR database. Pair-end reads of RNAseq were aligned to Human reference (hg38) using HISAT2. To identify the differentially mutated genes associated with patients with high/low MMAI scores, samples are binned separately based on the median or quartile of the MMAI score, and Fisher’s exact test is used to evaluate differentially mutated genes between bins. Differential expression of genes were evaluated using DEseq2 and gene set enrichment analysis. Uniform manifold approximation and projection (UMAP) was applied to 128 VFs from the MMAI score and correlated with mutations using Wilcoxon test.
Results: Median follow-up was 34.7 months. Patients with high MMAI scores (top quartile) had worse overall survival compared to those with low MMAI scores (bottom three quartiles, p=0.017). MMAI scores were significantly higher in synchronous than those with metachronous omCSPC (p<0.05). DNAseq showed high MMAI score (top median or quartile) had a trend towards more WNT pathway (APC/CTNNB1) mutations or had significantly more BRCA2/ATM mutations than those with low MMAI scores (p=0.13 and p=0.008, respectively). Low MMAI (<median) score patients had more SPOP mutations, which are known to correlate with better prognosis (p=0.03). RNAseq showed differentially expressed genes of patients with high MMAI scores were enriched for the EMT pathway (p<0.01). UMAP showed the 128 VFs of the MMAI algorithm formed 4 distinct clusters and a specific cluster correlated with WNT pathway mutations.
Conclusion: We found that the digital pathology MMAI algorithm developed-validated on localized CSPC is also prognostic in omCSPC. Furthermore, we found for the first time higher MMAI scores are correlated with mutations and transcriptional programs in metastatic pathways (e.g. WNT, BRCA2/ATM, EMT) in metachronous omCSPC. These data suggest that digital pathology-based MMAI algorithms identify phenotypic pathologic features that are representative of underlying biological genomic and transcriptomic processes of prognostic importance. Next steps are validation with spatial genomic methods.