Association of KRAS G12C Status with Age at Onset of Metastatic Colorectal Cancer.

The association of age at the onset of CRC and the prevalence of a KRAS G12C mutation is unclear. A retrospective, multicenter study evaluating metastatic CRC patients from January 2019 to July 2023, treated at the Oncoclinicas units and tested for tissue based KRAS/NRAS and BRAF mutations in a centralized genomics lab. A mismatch repair (MMR) status was retrieved from different labs and electronic medical records, as were patient demographics (age, gender) and tumor sidedness. The chi-square test was used to examine the association between clinical and molecular variables, with p value < 0.05 being statistically significant. A total of 858 cases were included. The median age was 63.7 years (range 22-95) and 17.4% were less than 50 years old at the diagnosis of metastatic CRC. Male patients represented 50.3% of the population. The sidedness distribution was as follows: left side 59.2%, right side 36.8% and not specified 4%. The prevalence of the KRAS mutation was 49.4% and the NRAS mutation was 3.9%. Among KRAS mutated tumors, the most common variants were G12V (27.6%) and G12D (23.5%), while KRAS G12C was less frequent (6.4%), which represented 3.1% of the overall population. The BRAF mutant cases were 7.3% and most commonly V600E. Only five (<1%) non-V600E mutations were detected. MSI-high or dMMR was present in 14 cases (1.6%). In the age-stratified analysis, left-sidedness (p < 0.001) and a KRAS G12C mutation (p = 0.046) were associated with a younger age (<50 years). In the sidedness-stratified analysis, a BRAF mutation (p = 0.001) and MSI-high/dMMR status (p = 0.009) were more common in right-sided tumors. Our data suggest that KRAS G12C mutations are more frequent in early-onset metastatic CRC. To the best of our knowledge, this is the largest cohort in the Latin American population with metastatic CRC reporting RAS, BRAF and MSI/MMR status.

Precision cancer medicine: Concepts, current practice, and future developments.

Precision cancer medicine is a multidisciplinary team effort that requires involvement and commitment of many stakeholders including the society at large. Building on the success of significant advances in precision therapy for oncological patients over the last two decades, future developments will be significantly shaped by improvements in scalable molecular diagnostics in which increasingly complex multilayered datasets require transformation into clinically useful information guiding patient management at fast turnaround times. Adaptive profiling strategies involving tissue- and liquid-based testing that account for the immense plasticity of cancer during the patient's journey and also include early detection approaches are already finding their way into clinical routine and will become paramount. A second major driver is the development of smart clinical trials and trial concepts which, complemented by real-world evidence, rapidly broaden the spectrum of therapeutic options. Tight coordination with regulatory agencies and health technology assessment bodies is crucial in this context. Multicentric networks operating nationally and internationally are key in implementing precision oncology in clinical practice and support developing and improving the ecosystem and framework needed to turn invocation into benefits for patients. The review provides an overview of the diagnostic tools, innovative clinical studies, and collaborative efforts needed to realize precision cancer medicine.

PCIG: a web-based application to explore immune-genomics interactions across cancer types.

MOTIVATION: Genomic alterations can modulate the tumor immunophenotype depending on their nature and tissue of origin. Although this immune-genomic interaction may shape disease progression and response to immunotherapy, the factors governing such dynamics and the influence of each tissue-specific context remain poorly understood. RESULTS: Here, we have developed the PanCancer ImmunoGenomics (PCIG) tool, a web-based resource that provides researchers with the opportunity to mine immunome-genome relationships across several cancer types using data from the Pan-Cancer Analysis of Whole-Genomes (PCAWG) study, which comprises >2,600 samples spanning across 20 different cancer primary sites. PCIG yields an integrative analysis of the crosstalk between somatic genomic alterations and different immune features, thus helping to understand immune response-related processes. AVAILABILITY AND IMPLEMENTATION: PCIG is freely available at https://pcig.vhio.net and is supported by all major web browsers. PCIG was developed with Django, which is a Python-based free and open-source framework, and it uses SQL Server as a relational database management system. The code is freely available for download at GitHub https://github.com/AnnaPG/PCIG and in its online supplementary material. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

A Comprehensive Biomarker Analysis of Microsatellite Unstable/Mismatch Repair Deficient Colorectal Cancer Cohort Treated with Immunotherapy.

The search for immunotherapy biomarkers in Microsatellite Instability High/Deficient Mismatch Repair system (MSI-H/dMMR) metastatic colorectal cancer (mCRC) is an unmet need. Sixteen patients with mCRC and MSI-H/dMMR (determined by either immunohistochemistry or polymerase chain reaction) treated with PD-1/PD-L1 inhibitors at our institution were included. According to whether the progression-free survival with PD-1/PD-L1 inhibitors was longer than 6 months or shorter, patients were clustered into the IT-responder group (n: 9 patients) or IT-resistant group (n: 7 patients), respectively. In order to evaluate determinants of benefit with PD-1/PD-L1 inhibitors, we performed multimodal analysis including genomics (through NGS panel tumour-only with 431 genes) and the immune microenvironment (using CD3, CD8, FOXP3 and PD-L1 antibodies). The following mutations were more frequent in IT-resistant compared with IT-responder groups: B2M (4/7 versus 2/9), CTNNB1 (2/7 versus 0/9), and biallelic PTEN (3/7 versus 1/9). Biallelic ARID1A mutations were found exclusively in the IT-responder group (4/9 patients). Tumour mutational burden did not correlate with immunotherapy benefit, neither the rate of indels in homopolymeric regions. Of note, biallelic ARID1A mutated tumours had the highest immune infiltration and PD-L1 scores, contrary to tumours with CTNNB1 mutation. Immune microenvironment analysis showed higher densities of different T cell subpopulations and PD-L1 expression in IT-responders. Misdiagnosis of MSI-H/dMMR inferred by discordances between immunohistochemistry and polymerase chain reaction was only found in the IT-resistant population (3/7 patients). Biallelic ARID1A mutations and Wnt signalling activation through CTNNB1 mutation were associated with high and low T cell immune infiltrates, respectively, and deserve special attention as determinants of response to PD-1/PD-L1 inhibitors. The non-MSI-H phenotype in dMMR is associated with poor benefit to immunotherapy. Our results suggest that mechanisms of resistance to immunotherapy are multi-factorial.

Perspectives on emerging technologies, personalised medicine, and clinical research for cancer control in Latin America and the Caribbean.

Challenges of health systems in Latin America and the Caribbean include accessibility, inequity, segmentation, and poverty. These challenges are similar in different countries of the region and transcend national borders. The increasing digital transformation of health care holds promise of more precise interventions, improved health outcomes, increased efficiency, and ultimately reduced health-care costs. In Latin America and the Caribbean, the adoption of digital health tools is in early stages and the quality of cancer registries, electronic health records, and structured databases are problematic. Cancer research and innovation in the region are limited due to inadequate academic resources and translational research is almost fully dependent on public funding. Regulatory complexity and extended timelines jeopardise the potential improvement in participation in international studies. Emerging technologies, artificial intelligence, big data, and cancer research represent an opportunity to address the health-care challenges in Latin America and the Caribbean collectively, by optimising national capacities, sharing and comparing best practices, and transferring scientific and technical capabilities.

PI3K activation promotes resistance to eribulin in HER2-negative breast cancer.

BACKGROUND: Eribulin is a microtubule-targeting agent approved for the treatment of advanced or metastatic breast cancer (BC) previously treated with anthracycline- and taxane-based regimens. PIK3CA mutation is associated with worse response to chemotherapy in oestrogen receptor-positive (ER+)/human epidermal growth factor receptor 2-negative (HER2-) metastatic BC. We aimed to evaluate the role of phosphoinositide 3-kinase (PI3K)/AKT pathway mutations in eribulin resistance. METHODS: Resistance to eribulin was evaluated in HER2- BC cell lines and patient-derived tumour xenografts, and correlated with a mutation in the PI3K/AKT pathway. RESULTS: Eleven out of 23 HER2- BC xenografts treated with eribulin exhibited disease progression. No correlation with ER status was detected. Among the resistant models, 64% carried mutations in PIK3CA, PIK3R1 or AKT1, but only 17% among the sensitive xenografts (P = 0.036). We observed that eribulin treatment induced AKT phosphorylation in vitro and in patient tumours. In agreement, the addition of PI3K inhibitors reversed primary and acquired resistance to eribulin in xenograft models, regardless of the genetic alterations in PI3K/AKT pathway or ER status. Mechanistically, PI3K blockade reduced p21 levels likely enabling apoptosis, thus sensitising to eribulin treatment. CONCLUSIONS: PI3K pathway activation induces primary resistance or early adaptation to eribulin, supporting the combination of PI3K inhibitors and eribulin for the treatment of HER2- BC patients.

A CT-based Radiomics Signature Is Associated with Response to Immune Checkpoint Inhibitors in Advanced Solid Tumors.

Background Reliable predictive imaging markers of response to immune checkpoint inhibitors are needed. Purpose To develop and validate a pretreatment CT-based radiomics signature to predict response to immune checkpoint inhibitors in advanced solid tumors. Materials and Methods In this retrospective study, a radiomics signature was developed in patients with advanced solid tumors (including breast, cervix, gastrointestinal) treated with anti-programmed cell death-1 or programmed cell death ligand-1 monotherapy from August 2012 to May 2018 (cohort 1). This was tested in patients with bladder and lung cancer (cohorts 2 and 3). Radiomics variables were extracted from all metastases delineated at pretreatment CT and selected by using an elastic-net model. A regression model combined radiomics and clinical variables with response as the end point. Biologic validation of the radiomics score with RNA profiling of cytotoxic cells (cohort 4) was assessed with Mann-Whitney analysis. Results The radiomics signature was developed in 85 patients (cohort 1: mean age, 58 years ± 13 [standard deviation]; 43 men) and tested on 46 patients (cohort 2: mean age, 70 years ± 12; 37 men) and 47 patients (cohort 3: mean age, 64 years ± 11; 40 men). Biologic validation was performed in a further cohort of 20 patients (cohort 4: mean age, 60 years ± 13; 14 men). The radiomics signature was associated with clinical response to immune checkpoint inhibitors (area under the curve [AUC], 0.70; 95% CI: 0.64, 0.77; P < .001). In cohorts 2 and 3, the AUC was 0.67 (95% CI: 0.58, 0.76) and 0.67 (95% CI: 0.56, 0.77; P < .001), respectively. A radiomics-clinical signature (including baseline albumin level and lymphocyte count) improved on radiomics-only performance (AUC, 0.74 [95% CI: 0.63, 0.84; P < .001]; Akaike information criterion, 107.00 and 109.90, respectively). Conclusion A pretreatment CT-based radiomics signature is associated with response to immune checkpoint inhibitors, likely reflecting the tumor immunophenotype. © RSNA, 2021 Online supplemental material is available for this article. See also the editorial by Summers in this issue.

Combination of novel molecular targeted agent plus R-CHOP-based regimen versus R-CHOP alone in previously untreated diffuse large B-cell lymphoma (DLBCL) patients: a systematic review and meta-analysis.

The addition of molecular targeted agents (MTAs) to R-CHOP has been one of the main focuses of research in patients with DLBCL. Despite encouraging preliminary results, recent randomized controlled trials (RCT) have not shown a definitive benefit over standard R-CHOP. Here we conducted a systematic review and meta-analysis to investigate the impact of this strategy. A systematic literature review was conducted to identify RCT that evaluated the addition of MTA to R-CHOP-based regimen versus R-CHOP alone in previously untreated DLBCL patients. Fixed and random effects models were used to estimate pooled hazard ratios (HRs) and odds ratios (ORs) with 95% confidence intervals (CI). Progression-free survival (PFS), overall survival, and adverse events (AE) were analyzed. A total of seven RCT including 3,255 patients with DLBCL met the eligibility criteria. Three different types of MTAs (bortezomib, ibrutinib, and lenalidomide) were investigated in combination with R-CHOP. Overall, R-CHOP plus MTA showed a slightly better PFS (HR=0.86; 95% CI: 0.76-0.98). No differences were observed according to the cell of origin subtype of DLBCL. Interestingly, patients younger than 60 years had a significantly better PFS with R-CHOP plus MTAs (HR=0.72; 95% CI: 0.56-0.93), while no benefit was observed in patients older than 60 years (HR=0.96). The combination strategy showed higher odds to develop serious AEs (OR= 1.46, 95% CI 1.11-1.91). R-CHOP plus MTA seems only to slightly improve PFS in patients with DLBCL, particularly in younger patients. An increase in toxicity was observed in comparison to R-CHOP.

Minimizing acquisition-related radiomics variability by image resampling and batch effect correction to allow for large-scale data analysis.

OBJECTIVE: To identify CT-acquisition parameters accounting for radiomics variability and to develop a post-acquisition CT-image correction method to reduce variability and improve radiomics classification in both phantom and clinical applications. METHODS: CT-acquisition protocols were prospectively tested in a phantom. The multi-centric retrospective clinical study included CT scans of patients with colorectal/renal cancer liver metastases. Ninety-three radiomics features of first order and texture were extracted. Intraclass correlation coefficients (ICCs) between CT-acquisition protocols were evaluated to define sources of variability. Voxel size, ComBat, and singular value decomposition (SVD) compensation methods were explored for reducing the radiomics variability. The number of robust features was compared before and after correction using two-proportion z test. The radiomics classification accuracy (K-means purity) was assessed before and after ComBat- and SVD-based correction. RESULTS: Fifty-three acquisition protocols in 13 tissue densities were analyzed. Ninety-seven liver metastases from 43 patients with CT from two vendors were included. Pixel size, reconstruction slice spacing, convolution kernel, and acquisition slice thickness are relevant sources of radiomics variability with a percentage of robust features lower than 80%. Resampling to isometric voxels increased the number of robust features when images were acquired with different pixel sizes (p < 0.05). SVD-based for thickness correction and ComBat correction for thickness and combined thickness-kernel increased the number of reproducible features (p < 0.05). ComBat showed the highest improvement of radiomics-based classification in both the phantom and clinical applications (K-means purity 65.98 vs 73.20). CONCLUSION: CT-image post-acquisition processing and radiomics normalization by means of batch effect correction allow for standardization of large-scale data analysis and improve the classification accuracy. KEY POINTS: • The voxel size (accounting for the pixel size and slice spacing), slice thickness, and convolution kernel are relevant sources of CT-radiomics variability. • Voxel size resampling increased the mean percentage of robust CT-radiomics features from 59.50 to 89.25% when comparing CT scans acquired with different pixel sizes and from 71.62 to 82.58% when the scans were acquired with different slice spacings. • ComBat batch effect correction reduced the CT-radiomics variability secondary to the slice thickness and convolution kernel, improving the capacity of CT-radiomics to differentiate tissues (in the phantom application) and the primary tumor type from liver metastases (in the clinical application).

The Evolving Role of Consensus Molecular Subtypes: a Step Beyond Inpatient Selection for Treatment of Colorectal Cancer.

OPINION STATEMENT: The heterogenous nature of colorectal cancer (CRC) renders it a major clinical challenge. Increasing genomic understanding of CRC has improved our knowledge of this heterogeneity and the main cancer drivers, with significant improvements in clinical outcomes. Comprehensive molecular characterization has allowed clinicians a more precise range of treatment options based on biomarker selection. Furthermore, this deep molecular understanding likely extends therapeutic options to a larger number of patients. The biological associations of consensus molecular subtypes (CMS) with clinical outcomes in localized CRC have been validated in retrospective clinical trials. The prognostic role of CMS has also been confirmed in the metastatic setting, with CMS2 having the best prognosis, whereas CMS1 tumors are associated with a higher risk of progression and death after chemotherapy. Similarly, according to mesenchymal features and immunosuppressive molecules, CMS1 responds to immunotherapy, whereas CMS4 has a poorer prognosis, suggesting that a CMS1 signature could identify patients who may benefit from immune checkpoint inhibitors regardless of microsatellite instability (MSI) status. The main goal of these comprehensive analyses is to switch from "one marker-one drug" to "multi-marker drug combinations" allowing oncologists to give "the right drug to the right patient." Despite the revealing data from transcriptomic analyses, the high rate of intra-tumoral heterogeneity across the different CMS subgroups limits its incorporation as a predictive biomarker. In clinical practice, when feasible, comprehensive genomic tests should be performed to identify potentially targetable alterations, particularly in RAS/BRAF wild-type, MSI, and right-sided tumors. Furthermore, CMS has not only been associated with clinical outcomes and specific tumor and patient phenotypes but also with specific microbiome patterns. Future steps will include the integration of clinical features, genomics, transcriptomics, and microbiota to select the most accurate biomarkers to identify optimal treatments, improving individual clinical outcomes. In summary, CMS is context specific, identifies a level of heterogeneity beyond standard genomic biomarkers, and offers a means of maximizing personalized therapy.

Molecular markers identify subtypes of stage III colon cancer associated with patient outcomes.

BACKGROUND & AIMS: Categorization of colon cancers into distinct subtypes using a combination of pathway-based biomarkers could provide insight into stage-independent variability in outcomes. METHODS: We used a polymerase chain reaction-based assay to detect mutations in BRAF (V600E) and in KRAS in 2720 stage III cancer samples, collected prospectively from patients participating in an adjuvant chemotherapy trial (NCCTG N0147). Tumors deficient or proficient in DNA mismatch repair (MMR) were identified based on detection of MLH1, MSH2, and MSH6 proteins and methylation of the MLH1 promoter. Findings were validated using tumor samples from a separate set of patients with stage III cancer (n = 783). Association with 5-year disease-free survival was evaluated using Cox proportional hazards models. RESULTS: Tumors were categorized into 5 subtypes based on MMR status and detection of BRAF or KRAS mutations which were mutually exclusive. Three subtypes were MMR proficient: those with mutations in BRAF (6.9% of samples), mutations in KRAS (35%), or tumors lacking either BRAF or KRAS mutations (49%). Two subtypes were MMR deficient: the sporadic type (6.8%) with BRAF mutation and/or or hypermethylation of MLH1 and the familial type (2.6%), which lacked BRAF(V600E) or hypermethylation of MLH1. A higher percentage of MMR-proficient tumors with BRAF(V600E) were proximal (76%), high-grade (44%), N2 stage (59%), and detected in women (59%), compared with MMR-proficient tumors without BRAF(V600E) or KRAS mutations (33%, 19%, 41%, and 42%, respectively; all P < .0001). A significantly lower proportion of patients with MMR-proficient tumors with mutant BRAF (hazard ratio = 1.43; 95% confidence interval: 1.11-1.85; Padjusted = .0065) or mutant KRAS (hazard ratio = 1.48; 95% confidence interval: 1.27-1.74; Padjusted < .0001) survived disease-free for 5 years compared with patients whose MMR-proficient tumors lacked mutations in either gene. Disease-free survival rates of patients with MMR-deficient sporadic or familial subtypes was similar to those of patients with MMR-proficient tumors without BRAF or KRAS mutations. The observed differences in survival rates of patients with different tumor subtypes were validated in an independent cohort. CONCLUSIONS: We identified subtypes of stage III colon cancer, based on detection of mutations in BRAF (V600E) or KRAS, and MMR status that show differences in clinical and pathologic features and disease-free survival. Patients with MMR-proficient tumors and BRAF or KRAS mutations had statistically shorter survival times than patients whose tumors lacked these mutations. The tumor subtype found in nearly half of the study cohort (MMR-proficient without BRAF(V600E) or KRAS mutations) had similar outcomes to those of patients with MMR-deficient cancers.

Precision medicine for urothelial carcinoma: An international perspective.

The treatment landscape of urothelial cancers has evolved in the last decade with the approval of chemotherapy, immune checkpoint inhibitors, targeted therapies, and antibody drug conjugates. Although improvements in response and survival have been achieved with these strategies, in some scenarios their benefit is still questionable. Current efforts to identify prognostic and predictive biomarkers are crucial for better patient selection and treatment outcomes. In this paper we will review the most promising biomarkers under investigation, such as molecular classifiers, genomic alterations, programmed cell death ligand 1 expression, tumor mutational burden, circulating tumor DNA, urinary biomarkers among others, for muscle invasive bladder cancer and metastatic urothelial cancers. Deeper understanding of these biomarkers will aid clinical decision-making and help tailor treatment strategies.

Independent transcriptional patterns reveal biological processes associated with disease-free survival in early colorectal cancer.

BACKGROUND: Bulk transcriptional profiles of early colorectal cancer (CRC) can fail to detect biological processes associated with disease-free survival (DFS) if the transcriptional patterns are subtle and/or obscured by other processes' patterns. Consensus-independent component analysis (c-ICA) can dissect such transcriptomes into statistically independent transcriptional components (TCs), capturing both pronounced and subtle biological processes. METHODS: In this study we (1) integrated transcriptomes (n = 4228) from multiple early CRC studies, (2) performed c-ICA to define the TC landscape within this integrated data set, 3) determined the biological processes captured by these TCs, (4) performed Cox regression to identify DFS-associated TCs, (5) performed random survival forest (RSF) analyses with activity of DFS-associated TCs as classifiers to identify subgroups of patients, and 6) performed a sensitivity analysis to determine the robustness of our results RESULTS: We identify 191 TCs, 43 of which are associated with DFS, revealing transcriptional diversity among DFS-associated biological processes. A prominent example is the epithelial-mesenchymal transition (EMT), for which we identify an association with nine independent DFS-associated TCs, each with coordinated upregulation or downregulation of various sets of genes. CONCLUSIONS: This finding indicates that early CRC may have nine distinct routes to achieve EMT, each requiring a specific peri-operative treatment strategy. Finally, we stratify patients into DFS patient subgroups with distinct transcriptional patterns associated with stage 2 and stage 3 CRC.

While treatments for patients with colorectal cancer have improved, many patients (around 30-50%) have cancers that will eventually relapse and these patients will die due to their disease. Researchers have been studying the genes involved in colorectal cancer to help us understand why some cancers might relapse. However, current methods to do this may miss subtle or hidden patterns in the gene activity related to cancer relapse. To deal with this, we used a special method called consensus-independent component analysis (c-ICA) to dig more deeply into the activity of genes. This helped us to uncover some potential biological processes underpinning colorectal cancer relapse, which ultimately could help researchers to identify better treatments for patients with colorectal cancer.

A whirl of radiomics-based biomarkers in cancer immunotherapy, why is large scale validation still lacking?

The search for understanding immunotherapy response has sparked interest in diverse areas of oncology, with artificial intelligence (AI) and radiomics emerging as promising tools, capable of gathering large amounts of information to identify suitable patients for treatment. The application of AI in radiology has grown, driven by the hypothesis that radiology images capture tumor phenotypes and thus could provide valuable insights into immunotherapy response likelihood. However, despite the rapid growth of studies, no algorithms in the field have reached clinical implementation, mainly due to the lack of standardized methods, hampering study comparisons and reproducibility across different datasets. In this review, we performed a comprehensive assessment of published data to identify sources of variability in radiomics study design that hinder the comparison of the different model performance and, therefore, clinical implementation. Subsequently, we conducted a use-case meta-analysis using homogenous studies to assess the overall performance of radiomics in estimating programmed death-ligand 1 (PD-L1) expression. Our findings indicate that, despite numerous attempts to predict immunotherapy response, only a limited number of studies share comparable methodologies and report sufficient data about cohorts and methods to be suitable for meta-analysis. Nevertheless, although only a few studies meet these criteria, their promising results underscore the importance of ongoing standardization and benchmarking efforts. This review highlights the importance of uniformity in study design and reporting. Such standardization is crucial to enable meaningful comparisons and demonstrate the validity of biomarkers across diverse populations, facilitating their implementation into the immunotherapy patient selection process.

Weakly Supervised Deep Learning Predicts Immunotherapy Response in Solid Tumors Based on PD-L1 Expression.

Programmed death-ligand 1 (PD-L1) IHC is the most commonly used biomarker for immunotherapy response. However, quantification of PD-L1 status in pathology slides is challenging. Neither manual quantification nor a computer-based mimicking of manual readouts is perfectly reproducible, and the predictive performance of both approaches regarding immunotherapy response is limited. In this study, we developed a deep learning (DL) method to predict PD-L1 status directly from raw IHC image data, without explicit intermediary steps such as cell detection or pigment quantification. We trained the weakly supervised model on PD-L1-stained slides from the non-small cell lung cancer (NSCLC)-Memorial Sloan Kettering (MSK) cohort (N = 233) and validated it on the pan-cancer-Vall d'Hebron Institute of Oncology (VHIO) cohort (N = 108). We also investigated the performance of the model to predict response to immune checkpoint inhibitors (ICI) in terms of progression-free survival. In the pan-cancer-VHIO cohort, the performance was compared with tumor proportion score (TPS) and combined positive score (CPS). The DL model showed good performance in predicting PD-L1 expression (TPS ≥ 1%) in both NSCLC-MSK and pan-cancer-VHIO cohort (AUC 0.88 ± 0.06 and 0.80 ± 0.03, respectively). The predicted PD-L1 status showed an improved association with response to ICIs [HR: 1.5 (95% confidence interval: 1-2.3), P = 0.049] compared with TPS [HR: 1.4 (0.96-2.2), P = 0.082] and CPS [HR: 1.2 (0.79-1.9), P = 0.386]. Notably, our explainability analysis showed that the model does not just look at the amount of brown pigment in the IHC slides, but also considers morphologic factors such as lymphocyte conglomerates. Overall, end-to-end weakly supervised DL shows potential for improving patient stratification for cancer immunotherapy by analyzing PD-L1 IHC, holistically integrating morphology and PD-L1 staining intensity. SIGNIFICANCE: The weakly supervised DL model to predict PD-L1 status from raw IHC data, integrating tumor staining intensity and morphology, enables enhanced patient stratification in cancer immunotherapy compared with traditional pathologist assessment.

Identification of Precise 3D CT Radiomics for Habitat Computation by Machine Learning in Cancer.

Purpose To identify precise three-dimensional radiomics features in CT images that enable computation of stable and biologically meaningful habitats with machine learning for cancer heterogeneity assessment. Materials and Methods This retrospective study included 2436 liver or lung lesions from 605 CT scans (November 2010-December 2021) in 331 patients with cancer (mean age, 64.5 years ± 10.1 [SD]; 185 male patients). Three-dimensional radiomics were computed from original and perturbed (simulated retest) images with different combinations of feature computation kernel radius and bin size. The lower 95% confidence limit (LCL) of the intraclass correlation coefficient (ICC) was used to measure repeatability and reproducibility. Precise features were identified by combining repeatability and reproducibility results (LCL of ICC ≥ 0.50). Habitats were obtained with Gaussian mixture models in original and perturbed data using precise radiomics features and compared with habitats obtained using all features. The Dice similarity coefficient (DSC) was used to assess habitat stability. Biologic correlates of CT habitats were explored in a case study, with a cohort of 13 patients with CT, multiparametric MRI, and tumor biopsies. Results Three-dimensional radiomics showed poor repeatability (LCL of ICC: median [IQR], 0.442 [0.312-0.516]) and poor reproducibility against kernel radius (LCL of ICC: median [IQR], 0.440 [0.33-0.526]) but excellent reproducibility against bin size (LCL of ICC: median [IQR], 0.929 [0.853-0.988]). Twenty-six radiomics features were precise, differing in lung and liver lesions. Habitats obtained with precise features (DSC: median [IQR], 0.601 [0.494-0.712] and 0.651 [0.52-0.784] for lung and liver lesions, respectively) were more stable than those obtained with all features (DSC: median [IQR], 0.532 [0.424-0.637] and 0.587 [0.465-0.703] for lung and liver lesions, respectively; P < .001). In the case study, CT habitats correlated quantitatively and qualitatively with heterogeneity observed in multiparametric MRI habitats and histology. Conclusion Precise three-dimensional radiomics features were identified on CT images that enabled tumor heterogeneity assessment through stable tumor habitat computation. Keywords: CT, Diffusion-weighted Imaging, Dynamic Contrast-enhanced MRI, MRI, Radiomics, Unsupervised Learning, Oncology, Liver, Lung Supplemental material is available for this article. © RSNA, 2024 See also the commentary by Sagreiya in this issue.