A proteogenomic portrait of lung squamous cell carcinoma.

Lung squamous cell carcinoma (LSCC) remains a leading cause of cancer death with few therapeutic options. We characterized the proteogenomic landscape of LSCC, providing a deeper exposition of LSCC biology with potential therapeutic implications. We identify NSD3 as an alternative driver in FGFR1-amplified tumors and low-p63 tumors overexpressing the therapeutic target survivin. SOX2 is considered undruggable, but our analyses provide rationale for exploring chromatin modifiers such as LSD1 and EZH2 to target SOX2-overexpressing tumors. Our data support complex regulation of metabolic pathways by crosstalk between post-translational modifications including ubiquitylation. Numerous immune-related proteogenomic observations suggest directions for further investigation. Proteogenomic dissection of CDKN2A mutations argue for more nuanced assessment of RB1 protein expression and phosphorylation before declaring CDK4/6 inhibition unsuccessful. Finally, triangulation between LSCC, LUAD, and HNSCC identified both unique and common therapeutic vulnerabilities. These observations and proteogenomics data resources may guide research into the biology and treatment of LSCC.

Geographic variation of mutagenic exposures in kidney cancer genomes.

International differences in the incidence of many cancer types indicate the existence of carcinogen exposures that have not yet been identified by conventional epidemiology make a substantial contribution to cancer burden1. In clear cell renal cell carcinoma, obesity, hypertension and tobacco smoking are risk factors, but they do not explain the geographical variation in its incidence2. Underlying causes can be inferred by sequencing the genomes of cancers from populations with different incidence rates and detecting differences in patterns of somatic mutations. Here we sequenced 962 clear cell renal cell carcinomas from 11 countries with varying incidence. The somatic mutation profiles differed between countries. In Romania, Serbia and Thailand, mutational signatures characteristic of aristolochic acid compounds were present in most cases, but these were rare elsewhere. In Japan, a mutational signature of unknown cause was found in more than 70% of cases but in less than 2% elsewhere. A further mutational signature of unknown cause was ubiquitous but exhibited higher mutation loads in countries with higher incidence rates of kidney cancer. Known signatures of tobacco smoking correlated with tobacco consumption, but no signature was associated with obesity or hypertension, suggesting that non-mutagenic mechanisms of action underlie these risk factors. The results of this study indicate the existence of multiple, geographically variable, mutagenic exposures that potentially affect tens of millions of people and illustrate the opportunities for new insights into cancer causation through large-scale global cancer genomics.

Mapping clustered mutations in cancer reveals APOBEC3 mutagenesis of ecDNA.

Clustered somatic mutations are common in cancer genomes and previous analyses reveal several types of clustered single-base substitutions, which include doublet- and multi-base substitutions1-5, diffuse hypermutation termed omikli6, and longer strand-coordinated events termed kataegis3,7-9. Here we provide a comprehensive characterization of clustered substitutions and clustered small insertions and deletions (indels) across 2,583 whole-genome-sequenced cancers from 30 types of cancer10. Clustered mutations were highly enriched in driver genes and associated with differential gene expression and changes in overall survival. Several distinct mutational processes gave rise to clustered indels, including signatures that were enriched in tobacco smokers and homologous-recombination-deficient cancers. Doublet-base substitutions were caused by at least 12 mutational processes, whereas most multi-base substitutions were generated by either tobacco smoking or exposure to ultraviolet light. Omikli events, which have previously been attributed to APOBEC3 activity6, accounted for a large proportion of clustered substitutions; however, only 16.2% of omikli matched APOBEC3 patterns. Kataegis was generated by multiple mutational processes, and 76.1% of all kataegic events exhibited mutational patterns that are associated with the activation-induced deaminase (AID) and APOBEC3 family of deaminases. Co-occurrence of APOBEC3 kataegis and extrachromosomal DNA (ecDNA), termed kyklonas (Greek for cyclone), was found in 31% of samples with ecDNA. Multiple distinct kyklonic events were observed on most mutated ecDNA. ecDNA containing known cancer genes exhibited both positive selection and kyklonic hypermutation. Our results reveal the diversity of clustered mutational processes in human cancer and the role of APOBEC3 in recurrently mutating and fuelling the evolution of ecDNA.

Mechanisms of APOBEC3 mutagenesis in human cancer cells.

The APOBEC3 family of cytosine deaminases has been implicated in some of the most prevalent mutational signatures in cancer1-3. However, a causal link between endogenous APOBEC3 enzymes and mutational signatures in human cancer genomes has not been established, leaving the mechanisms of APOBEC3 mutagenesis poorly understood. Here, to investigate the mechanisms of APOBEC3 mutagenesis, we deleted implicated genes from human cancer cell lines that naturally generate APOBEC3-associated mutational signatures over time4. Analysis of non-clustered and clustered signatures across whole-genome sequences from 251 breast, bladder and lymphoma cancer cell line clones revealed that APOBEC3A deletion diminished APOBEC3-associated mutational signatures. Deletion of both APOBEC3A and APOBEC3B further decreased APOBEC3 mutation burdens, without eliminating them. Deletion of APOBEC3B increased APOBEC3A protein levels, activity and APOBEC3A-mediated mutagenesis in some cell lines. The uracil glycosylase UNG was required for APOBEC3-mediated transversions, whereas the loss of the translesion polymerase REV1 decreased overall mutation burdens. Together, these data represent direct evidence that endogenous APOBEC3 deaminases generate prevalent mutational signatures in human cancer cells. Our results identify APOBEC3A as the main driver of these mutations, indicate that APOBEC3B can restrain APOBEC3A-dependent mutagenesis while contributing its own smaller mutation burdens and dissect mechanisms that translate APOBEC3 activities into distinct mutational signatures.

The repertoire of mutational signatures in human cancer.

Somatic mutations in cancer genomes are caused by multiple mutational processes, each of which generates a characteristic mutational signature1. Here, as part of the Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium2 of the International Cancer Genome Consortium (ICGC) and The Cancer Genome Atlas (TCGA), we characterized mutational signatures using 84,729,690 somatic mutations from 4,645 whole-genome and 19,184 exome sequences that encompass most types of cancer. We identified 49 single-base-substitution, 11 doublet-base-substitution, 4 clustered-base-substitution and 17 small insertion-and-deletion signatures. The substantial size of our dataset, compared with previous analyses3-15, enabled the discovery of new signatures, the separation of overlapping signatures and the decomposition of signatures into components that may represent associated-but distinct-DNA damage, repair and/or replication mechanisms. By estimating the contribution of each signature to the mutational catalogues of individual cancer genomes, we revealed associations of signatures to exogenous or endogenous exposures, as well as to defective DNA-maintenance processes. However, many signatures are of unknown cause. This analysis provides a systematic perspective on the repertoire of mutational processes that contribute to the development of human cancer.

Visualizing and exploring patterns of large mutational events with SigProfilerMatrixGenerator.

BACKGROUND: All cancers harbor somatic mutations in their genomes. In principle, mutations affecting between one and fifty base pairs are generally classified as small mutational events. Conversely, large mutational events affect more than fifty base pairs, and, in most cases, they encompass copy-number and structural variants affecting many thousands of base pairs. Prior studies have demonstrated that examining patterns of somatic mutations can be leveraged to provide both biological and clinical insights, thus, resulting in an extensive repertoire of tools for evaluating small mutational events. Recently, classification schemas for examining large-scale mutational events have emerged and shown their utility across the spectrum of human cancers. However, there has been no computationally efficient bioinformatics tool that allows visualizing and exploring these large-scale mutational events. RESULTS: Here, we present a new version of SigProfilerMatrixGenerator that now delivers integrated capabilities for examining large mutational events. The tool provides support for examining copy-number variants and structural variants under two previously developed classification schemas and it supports data from numerous algorithms and data modalities. SigProfilerMatrixGenerator is written in Python with an R wrapper package provided for users that prefer working in an R environment. CONCLUSIONS: The new version of SigProfilerMatrixGenerator provides the first standardized bioinformatics tool for optimized exploration and visualization of two previously developed classification schemas for copy number and structural variants. The tool is freely available at https://github.com/AlexandrovLab/SigProfilerMatrixGenerator with an extensive documentation at https://osf.io/s93d5/wiki/home/ .

Examining clustered somatic mutations with SigProfilerClusters.

MOTIVATION: Clustered mutations are found in the human germline as well as in the genomes of cancer and normal somatic cells. Clustered events can be imprinted by a multitude of mutational processes, and they have been implicated in both cancer evolution and development disorders. Existing tools for identifying clustered mutations have been optimized for a particular subtype of clustered event and, in most cases, relied on a predefined inter-mutational distance (IMD) cutoff combined with a piecewise linear regression analysis. RESULTS: Here, we present SigProfilerClusters, an automated tool for detecting all types of clustered mutations by calculating a sample-dependent IMD threshold using a simulated background model that takes into account extended sequence context, transcriptional strand asymmetries and regional mutation densities. SigProfilerClusters disentangles all types of clustered events from non-clustered mutations and annotates each clustered event into an established subclass, including the widely used classes of doublet-base substitutions, multi-base substitutions, omikli and kataegis. SigProfilerClusters outputs non-clustered mutations and clustered events using standard data formats as well as provides multiple visualizations for exploring the distributions and patterns of clustered mutations across the genome. AVAILABILITY AND IMPLEMENTATION: SigProfilerClusters is supported across most operating systems and made freely available at https://github.com/AlexandrovLab/SigProfilerClusters with an extensive documentation located at https://osf.io/qpmzw/wiki/home/. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Automating UbiFast for High-throughput and Multiplexed Ubiquitin Enrichment.

Robust methods for deep-scale enrichment and site-specific identification of ubiquitylation sites are necessary for characterizing the myriad roles of protein ubiquitylation. To this end we previously developed UbiFast, a sensitive method for highly multiplexed ubiquitylation profiling where K-ϵ-GG peptides are enriched with anti-K-ε-GG antibody and labeled on-antibody with isobaric labeling reagents for sample multiplexing. Here, we present robotic automation of the UbiFast method using a magnetic bead-conjugated K-ε-GG antibody (mK-ε-GG) and a magnetic particle processor. We report the identification of ∼20,000 ubiquitylation sites from a TMT10-plex with 500 µg input per sample processed in ∼2 h. Automation of the UbiFast method greatly increased the number of identified and quantified ubiquitylation sites, improved reproducibility, and significantly reduced processing time. The automated method also significantly reduced variability across process replicates compared with the manual method. The workflow enables processing of up to 96 samples in a single day making it suitable to study ubiquitylation in large sample sets. Here we demonstrate the applicability of the method to profile small amounts of tissue using breast cancer patient-derived xenograft (PDX) tissue samples.

Perceived problems with academic achievement in school-aged children with recurrent pain - a longitudinal study.

Aims: The overall aim was to explore the relationship between recurrent pain and perceived problems with academic achievement among boys and girls in middle and late elementary school. Methods: This 3-year follow-up study was based on data from the Study of Health in School-aged Children from Umeå (Sweden) and included children attending grade 6 in years 2003 and 2006, and a follow-up 3 years later in grade 9 (n = 1524, participation rate 90%). Results: Recurrent pain (head, stomach or back) at least doubled the odds of concurrent- and subsequent perceived problems with academic achievement. This applied for pain on a monthly and weekly basis, from single and multiple sites, and from each of the three studied pain sites. The odds increased with increasing pain frequency and number of pain sites. Problems with sleep, concentration or school absenteeism did not explain the association. Conclusions: Recurrent pain seems to be a potential predictor of perceived problems with academic achievement for school-aged children. This emphasises the importance of early identification and prevention of recurrent pain problems.

Generating realistic null hypothesis of cancer mutational landscapes using SigProfilerSimulator.

BACKGROUND: Performing a statistical test requires a null hypothesis. In cancer genomics, a key challenge is the fast generation of accurate somatic mutational landscapes that can be used as a realistic null hypothesis for making biological discoveries. RESULTS: Here we present SigProfilerSimulator, a powerful tool that is capable of simulating the mutational landscapes of thousands of cancer genomes at different resolutions within seconds. Applying SigProfilerSimulator to 2144 whole-genome sequenced cancers reveals: (i) that most doublet base substitutions are not due to two adjacent single base substitutions but likely occur as single genomic events; (ii) that an extended sequencing context of ± 2 bp is required to more completely capture the patterns of substitution mutational signatures in human cancer; (iii) information on false-positive discovery rate of commonly used bioinformatics tools for detecting driver genes. CONCLUSIONS: SigProfilerSimulator's breadth of features allows one to construct a tailored null hypothesis and use it for evaluating the accuracy of other bioinformatics tools or for downstream statistical analysis for biological discoveries. SigProfilerSimulator is freely available at https://github.com/AlexandrovLab/SigProfilerSimulator with an extensive documentation at https://osf.io/usxjz/wiki/home/ .

SigProfilerMatrixGenerator: a tool for visualizing and exploring patterns of small mutational events.

BACKGROUND: Cancer genomes are peppered with somatic mutations imprinted by different mutational processes. The mutational pattern of a cancer genome can be used to identify and understand the etiology of the underlying mutational processes. A plethora of prior research has focused on examining mutational signatures and mutational patterns from single base substitutions and their immediate sequencing context. We recently demonstrated that further classification of small mutational events (including substitutions, insertions, deletions, and doublet substitutions) can be used to provide a deeper understanding of the mutational processes that have molded a cancer genome. However, there has been no standard tool that allows fast, accurate, and comprehensive classification for all types of small mutational events. RESULTS: Here, we present SigProfilerMatrixGenerator, a computational tool designed for optimized exploration and visualization of mutational patterns for all types of small mutational events. SigProfilerMatrixGenerator is written in Python with an R wrapper package provided for users that prefer working in an R environment. SigProfilerMatrixGenerator produces fourteen distinct matrices by considering transcriptional strand bias of individual events and by incorporating distinct classifications for single base substitutions, doublet base substitutions, and small insertions and deletions. While the tool provides a comprehensive classification of mutations, SigProfilerMatrixGenerator is also faster and more memory efficient than existing tools that generate only a single matrix. CONCLUSIONS: SigProfilerMatrixGenerator provides a standardized method for classifying small mutational events that is both efficient and scalable to large datasets. In addition to extending the classification of single base substitutions, the tool is the first to provide support for classifying doublet base substitutions and small insertions and deletions. SigProfilerMatrixGenerator is freely available at https://github.com/AlexandrovLab/SigProfilerMatrixGenerator with an extensive documentation at https://osf.io/s93d5/wiki/home/ .

Low back pain and associated disability in Swedish adolescents.

INTRODUCTION: Low back pain (LBP) defined as ache or pain in the lowest part of the back is a common experience among people all over the world. The lifetime prevalence is reported to be as high as 84%, and the prevalence of LBP seems to be almost the same among adolescents as among adults. The risk for having LBP later in life if you experienced LBP in adolescence is high. MATERIAL AND METHODS: In this cross-sectional study of 2550 students aged 16-20 years, we used the Standardized Nordic Questionnaire for the analysis of musculo-skeletal symptoms. We studied gender difference, prevalence, and disability of Low back pain. We also studied differences in LBP in adolescent athletes depending on hours spent on sports or physical activity. RESULTS: Significantly, more girls than boys had had problems sometimes during their life. Those who participated in sports reported LBP to a significantly higher extent than those who were physically inactive in their spare time. Gender and spare time sports were important risk factors for getting LBP some time in life. There was a higher risk for girls to have low back pain problems more than 30 days or daily the last year if they had had low back pain some time earlier in life. CONCLUSION: This study shows that low back pain (LBP) is common among Swedish adolescents, more common among girls than boys. High sport activity was associated with the risk of having LBP, length of time with LBP, and disability due to LBP.

Self-rated health, subjective social status in school and socioeconomic status in adolescents: a cross-sectional study.

BACKGROUND: Social position, traditionally measured by objective data on socioeconomic status (SES), is linked to health status in adults. In adolescents, the association is more uncertain and there are some studies suggesting that subjective social status (SSS) might be more adequate in relation to health. This study aimed to examine associations between SSS in school, SES and self-rated health (SRH) in adolescent boys and girls. METHODS: A descriptive cross-sectional research design with quantitative survey data was used. The study involved 705 Swedish adolescents in upper secondary school (17-18-year-olds). SRH was measured with a single-item question and SSS by a question where adolescents were asked to assess their social position within their school. Formal education level of the parents was used as a proxy for objective SES. Univariable and multivariable ordinal regression analyses were conducted to assess the associations between SRH and SSS in school and SES. RESULTS: In the multivariable analysis, SSS in school was positively associated with SRH, whereas no significant association between SES and SRH was found. The proportion of adolescents with high SRH increased with higher steps on the SSS ladder. Significant gender differences were found in that boys rated their SRH and SSS in school higher than girls did. CONCLUSIONS: The study shows that self-rated health in adolescents is related to perceived social position in school. Subjective social status in school seems to be a useful health-related measure of social position in adolescents.

School experiences in relation to emotional and conduct problems in adolescence: a 3-year follow up study.

BACKGROUND: Mental health in adolescents has become a major public health issue. This study examined school experiences in relation to mental health (emotional problems and conduct problems) from early to middle adolescence. METHODS: This longitudinal 3-year follow up study used data from the Swedish Study of Health in School Children in Umeå. Analyses were conducted in 1379 participants that were attending grade six in 2003 or 2006 (age 12 years). KIDSCREEN-52 was used to assess school experiences and the Strengths and Difficulties Questionnaire for emotional and conduct problems. Statistical analyses included repeated measures ANOVA and multiple linear regressions. RESULTS: Positive school experiences decreased while emotional and conduct problem scores increased from grades six to nine. Positive school experiences were negatively associated with emotional and conduct problem scores and contributed to the explanation of mental health scores in middle adolescence after controlling for background factors. When baseline mental health problem scores were taken into account the association with early school experiences disappeared (except for conduct problems in boys). However, incorporating concurrent school experiences in the analysis increased the levels of explanation for emotional and conduct problem scores further. CONCLUSIONS: The results of this study confirm that school experiences are linked to emotional and conduct problems. That link may be stronger for conduct problems. In addition, the association of school experiences in early adolescence with later mental health may be overridden by concurrent school experiences in middle adolescence.

Mental problems and their socio-demographic determinants in young schoolchildren in Sweden, a country with high gender and income equality.

AIMS: Mental problems and their potential socio-demographic determinants were investigated in young schoolchildren in Sweden, a high-income country in the top of income- and gender-equality rankings. METHODS: Cross-sectional study of 1465 schoolchildren in grades 3 and 6. Mental health was measured by the Child Behavior Checklist and the Youth Self Report (Total problems and 14 specific problem areas). Potential socio-demographic determinants were sex, parental education and occupation, family structure, and immigrant status. RESULTS: Mental problems were present in 14% of the sixth graders and in 7% of the third graders. In grade 3, the mean total problem score was lower in girls than in boys, but the prevalence of problems at a subclinical/clinical level did not differ by sex. Furthermore, in nine to 13 of the 14 specific problem areas, problems were equally distributed by sex, parental education, parental occupation, immigrant status, and family structure. In grade 6, both the total mean score and the overall odds of subclinical/clinical problems were similar in girls and boys. Likewise, in all the specific problem areas, problems were evenly distributed by parental education and occupation, and only independently associated with immigrant status and family structure in one problem area. In five specific problem areas, boys had higher odds of problems than girls. CONCLUSIONS: This study shows that also in a relatively wealthy and equal country such as Sweden, mental problems are a significant child public health issue. The association between socio-demographic background and mental problems seems to be rather weak, but differ dependent on the type of mental problem in focus.

School experiences may be important determinants of mental health problems in middle childhood - a Swedish longitudinal population-based study.

AIM: Little is known about the association between school experiences and mental health in young schoolchildren. This study explored the cross-sectional and prospective associations between children's school experiences and mental health in middle childhood. METHODS: We gathered comprehensive population-based data on the school experiences and mental health of 592 schoolchildren attending grades three and six in Sweden (ages approximately nine and 12 years). The KIDSCREEN questionnaire was used to measure school experiences in both age groups while the Child Behavior Checklist and the Strengths and Difficulties Questionnaire measured mental health in grades three and six, respectively. RESULTS: Children with problematic school experiences in grade three had an approximately two times higher odds for concurrent total, internalised, externalised, attention-hyperactivity and social problems. They also had a 1.5-2.5 higher odds for these mental health problems three years later. Likewise, there was an association between problematic school experiences in grade three and lower levels of prosocial behaviour three years later. These associations were shown in both boys and girls, but were particularly pronounced in girls. CONCLUSION: This study indicated that school experiences in young schoolchildren may be important determinants of concurrent and later mental health problems.

Predictors of smoking among Swedish adolescents.

BACKGROUND: Smoking most often starts in adolescence, implying that understanding of predicting factors for smoking initiation during this time period is essential for successful smoking prevention. The aim of this study was to examine predicting factors in early adolescence for smoking in late adolescence. METHODS: Longitudinal cohort study, involving 649 Swedish adolescents from lower secondary school (12-13 years old) to upper secondary school (17-18 years old). Tobacco habits, behavioural, intra- and interpersonal factors and socio-demographic variables were assessed through questionnaires. Descriptive statistics, univariable and multivariable logistic regression were used to identify predicting factors. RESULTS: Smoking prevalence increased from 3.3% among 12-13 year olds to 25.1% among 17-18 year olds. Possible predictors of smoking were: female sex, lower parental education, poorer family mood, poorer self-rated health, poorer self-esteem, less negative attitude towards smoking, binge drinking, snus use and smoking. In a multivariable logistic regression analysis, female sex (OR 1.64, CI 1.08-2.49), medium and low self-esteem (medium: OR 1.57, CI 1.03-2.38, low: 2.79, CI 1.46-5.33), less negative attitude towards smoking (OR 2.81, CI 1.70-4.66) and ever using snus (OR 3.43, CI 1.78-6.62) remained significant independent predicting factors. CONCLUSIONS: The study stresses the importance of strengthening adolescents' self-esteem, promoting anti-smoking attitudes in early adolescence, as well as avoidance of early initiation of snus. Such measures should be joint efforts involving parents, schools, youth associations, and legislating authorities.

Adolescent self-reported health in relation to school factors: a multilevel analysis.

The aim of the study was to examine school-related determinants of self-reported health among adolescents. Questionnaire survey data comprising 4,972 students, Grades 7 through 9, from 20 schools in northern Sweden were used. Also, complimentary data about each school were collected from the Swedish National Agency for Education. Using multilevel logistic regression analyses, results showed that most variation in self-reported health was explained by individual-level differences. Truancy, bullying, and poor relations with teachers significantly increased the odds ratio of reporting poor general health, for boys and for girls. Most variables at the school level, for example, school size and student-teacher ratio, did not render significant associations with students' self-reported health. In conclusion, this study indicates that health promotion at school, including school health services, may benefit from focusing primarily on individual-level determinants of health, that is, students' relations to peers and teachers, without ignoring that bullying and weak student-teacher relationships also may induce school-level interventions.

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

PURPOSE: Cancers with homologous recombination deficiency (HRD) can benefit from platinum salts and poly(ADP-ribose) polymerase inhibitors. Standard diagnostic tests for detecting HRD require molecular profiling, which is not universally available. METHODS: We trained DeepHRD, a deep learning platform for predicting HRD from hematoxylin and eosin (H&E)-stained histopathological slides, using primary breast (n = 1,008) and ovarian (n = 459) cancers from The Cancer Genome Atlas (TCGA). DeepHRD was compared with four standard HRD molecular tests using breast (n = 349) and ovarian (n = 141) cancers from multiple independent data sets, including platinum-treated clinical cohorts with RECIST progression-free survival (PFS), complete response (CR), and overall survival (OS) endpoints. RESULTS: DeepHRD predicted HRD from held-out H&E-stained breast cancer slides in TCGA with an AUC of 0.81 (95% CI, 0.77 to 0.85). This performance was confirmed in two independent primary breast cancer cohorts (AUC, 0.76 [95% CI, 0.71 to 0.82]). In an external platinum-treated metastatic breast cancer cohort, samples predicted as HRD had higher complete CR (AUC, 0.76 [95% CI, 0.54 to 0.93]) with 3.7-fold increase in median PFS (14.4 v 3.9 months; P = .0019) and hazard ratio (HR) of 0.45 (P = .0047). There were no significant differences in nonplatinum treatment outcome by predicted HRD status in three breast cancer cohorts, including CR (AUC, 0.39) and PFS (HR, 0.98, P = .95) in taxane-treated metastatic breast cancer. Through transfer learning to high-grade serous ovarian cancer, DeepHRD-predicted HRD samples had better OS after first-line (HR, 0.46; P = .030) and neoadjuvant (HR, 0.49; P = .015) platinum therapy in two cohorts. CONCLUSION: DeepHRD can predict HRD in breast and ovarian cancers directly from routine H&E slides across multiple external cohorts, slide scanners, and tissue fixation variables. When compared with molecular testing, DeepHRD classified 1.8- to 3.1-fold more patients with HRD, which exhibited better OS in high-grade serous ovarian cancer and platinum-specific PFS in metastatic breast cancer.