Abstract:Abstract Estimating the status of homologous recombination deficiency (HRD) is of clinical interest because the homologous recombination-deficient cells are reported sensitive to poly-ADP ribose polymerase (PARP) inhibitors. As HRD status becomes a critical indicator, many approaches have been proposed to measure the HRD status, among which estimating HRD status from cancer sequencing data is the popular strategy, where the representative methods include HRDetect, SigMA, FoundationFocusCDx, MyChoiceHRD, etc. Although sequencing data is the same input of these bioinformatics tools, algorithms prefer different aspects. Some methods focus on germline variants in BRCA1/2 genes and somatic mutational signatures, while some others capture the genomic scars, including loss of heterozygosity (LOH), telomere allele imbalance (TAI) and large-scale state transition (LST). However, there are two major controversial points when applying these tools on clinical sequencing data. First, panel sequencing is the most popular sequencing plan in clinic, but gene panels usually carry only hundreds to a thousand genes. Thus, the panel sequencing can neither capture a large number of mutations to calculate mutational signatures, nor obtain large genomic scars, such as TAI and LST. Second, the associations between the mutations and HRD status is still unclear. Thus, scoring mutations and genomic scars is not reasonable. To solving these issues, we here proposed a machine learning-based method, called tmHRD, which estimates HRD status according to limited mutations and LOHs captured by clinical panel sequencing data. It consists of a semi-supervised learning (TRI) model and a multi-instance learning (MIL) model. The TRI model filters the false positives from mutation calls. Then, the filtered mutation calls from one patient is defined as a package, where each mutation call is an instance in the package. When training the MIL model, in each iteration, a package with known HRD status (label) is given to MIL model. MIL model calculates the correlations among instances and assigns an adjusted weight for each. Thus, the HRD status is only contributed by those mutations associated with. We tested tmHRD on several datasets obtained by the clinical trials we participated. Compared to the existing methods, tmHRD achieved better performance. The average recognition rate of tmHRD was approaching 90%, while the comparison methods show around 80%. The software package, tmHRD, is freely available at https://github.com/Sherwin-xjtu/TriMILhrd/ for academic usages only. Citation Format: Xuwen Wang, Ying Xu, Yanfang Guan, Xin Yi, Jiayin Wang. tmHRD: Accurately estimating the homologous recombination deficiency status from clinical panel sequencing data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr LB543.

expHRD: an individualized, transcriptome-based prediction model for homologous recombination deficiency assessment in cancer

IdentifiHR: predicting homologous recombination deficiency in high-grade serous ovarian carcinoma using gene expression

Prediction of homologous recombination deficiency from Oncomine Comprehensive Assay Plus correlating with SOPHiA DDM HRD Solution

AcornHRD: A novel tool to identify homologous recombination deficiency events in whole genome sequencing data.

Detecting HRD in whole-genome and whole-exome sequenced breast and ovarian cancers

DirectHRD enables sensitive scar-based classification of homologous recombination deficiency (HRD)

Abstract 2328: Next-generation sequencing-based regression algorithm to determine homologous recombination deficiency scores in a pan-cancer cohort

Genomic Scar Score: A Robust Model Predicting Homologous Recombination Deficiency Based on Genomic Instability

Optimized Detection of Homologous Recombination Deficiency Improves the Prediction of Clinical Outcomes in Cancer

HRD-MILN: Accurately Estimate Tumor Homologous Recombination Deficiency Status from Targeted Panel Sequencing Data

Homologous recombination deficiency in diverse cancer types and its correlation with platinum chemotherapy efficiency in ovarian cancer

Patient Assessment and Therapy Planning Based on Homologous Recombination Repair Deficiency

Abstract LB543: Tmhrd: Accurately Estimating the Homologous Recombination Deficiency Status from Clinical Panel Sequencing Data

Homologous Recombination Deficiency Score for Predicting Efficacy of Platinum-Based Chemotherapy and PARPi Maintenance Therapy in Ovarian Cancer.

Deep Learning Artificial Intelligence Predicts Homologous Recombination Deficiency and Platinum Response From Histologic Slides

Radiogenomics of homologous recombination deficiency (HRD) in breast cancer: Radiomics-based image biomarker for HRD.

Pan-cancer Analysis of Homologous Recombination Deficiency in Cell Lines

Abstract B031: Evaluating the consequence of homologous recombination deficiency in models of pediatric solid tumor

Modification of Homologous Recombination Deficiency Score Threshold and Association with Long-Term Survival in Epithelial Ovarian Cancer

Correlation of Homologous Recombination Deficiency in Head and Neck Cancer with Sensitivity to Parp Inhibition

A Risk Model Developed based on Homologous Recombination Deficiency Genes for Evaluating the Drug Sensitivity and Prognostic Prediction of Lung Adenocarcinoma