Deep learning-based prediction of molecular cancer biomarkers from tissue slides: A new tool for precision oncology

Training immunophenotyping deep learning models with the same-section ground truth cell label derivation method improves virtual staining accuracy

Introduction

Deep learning (DL) models predicting biomarker expression in images of hematoxylin and eosin (H&E)-stained tissues can improve access to multi-marker immunophenotyping, crucial for therapeutic monitoring, biomarker discovery, and personalized treatment development. Conventionally, these models are trained on ground truth cell labels derived from IHC-stained tissue sections adjacent to H&E-stained ones, which might be less accurate than labels from the same section. Although many such DL models have been developed, the impact of ground truth cell label derivation methods on their performance has not been studied.

Methodology

In this study, we assess the impact of cell label derivation on H&E model performance, with CD3+ T-cells in lung cancer tissues as a proof-of-concept. We compare two Pix2Pix generative adversarial network (P2P-GAN)-based virtual staining models: one trained with cell labels obtained from the same tissue section as the H&E-stained section (the 'same-section' model) and one trained on cell labels from an adjacent tissue section (the 'serial-section' model).

Results

We show that the same-section model exhibited significantly improved prediction performance compared to the 'serial-section' model. Furthermore, the same-section model outperformed the serial-section model in stratifying lung cancer patients within a public lung cancer cohort based on survival outcomes, demonstrating its potential clinical utility.

Discussion

Collectively, our findings suggest that employing ground truth cell labels obtained through the same-section approach boosts immunophenotyping DL solutions.

High-content microscopy reveals a morphological signature of bortezomib resistance

Drug resistance is a challenge in anticancer therapy. In many cases, cancers can be resistant to the drug prior to exposure, that is, possess intrinsic drug resistance. However, we lack target-independent methods to anticipate resistance in cancer cell lines or characterize intrinsic drug resistance without a priori knowledge of its cause. We hypothesized that cell morphology could provide an unbiased readout of drug resistance. To test this hypothesis, we used HCT116 cells, a mismatch repair-deficient cancer cell line, to isolate clones that were resistant or sensitive to bortezomib, a well-characterized proteasome inhibitor and anticancer drug to which many cancer cells possess intrinsic resistance. We then expanded these clones and measured high-dimensional single-cell morphology profiles using Cell Painting, a high-content microscopy assay. Our imaging- and computation-based profiling pipeline identified morphological features that differed between resistant and sensitive cells. We used these features to generate a morphological signature of bortezomib resistance. We then employed this morphological signature to analyze a set of HCT116 clones (five resistant and five sensitive) that had not been included in the signature training dataset, and correctly predicted sensitivity to bortezomib in seven cases, in the absence of drug treatment. This signature predicted bortezomib resistance better than resistance to other drugs targeting the ubiquitin-proteasome system, indicating specificity for mechanisms of resistance to bortezomib. Our results establish a proof-of-concept framework for the unbiased analysis of drug resistance using high-content microscopy of cancer cells, in the absence of drug treatment.

High-content microscopy reveals a morphological signature of bortezomib resistance

Drug resistance is a challenge in anticancer therapy, particularly with targeted therapeutics and cytotoxic compounds. In many cases, cancers can be resistant to the drug prior to exposure, i.e., possess intrinsic drug resistance. However, we lack target-independent methods to anticipate resistance in cancer cell lines or characterize intrinsic drug resistance without a priori knowledge of its cause. We hypothesized that cell morphology could provide an unbiased readout of drug sensitivity prior to treatment. We therefore isolated clonal cell lines that were either sensitive or resistant to bortezomib, a well-characterized proteasome inhibitor and anticancer drug to which many cancer cells possess intrinsic resistance. We then measured high-dimensional single-cell morphology profiles using Cell Painting, a high-content microscopy assay. Our imaging- and computation-based profiling pipeline identified morphological features typically different between resistant and sensitive clones. These features were compiled to generate a morphological signature of bortezomib resistance, which correctly predicted the bortezomib treatment response in seven of ten cell lines not included in the training dataset. This signature of resistance was specific to bortezomib over other drugs targeting the ubiquitin-proteasome system. Our results provide evidence that intrinsic morphological features of drug resistance exist and establish a framework for their identification.

Hepatic arterial infusion chemotherapy and immune checkpoint inhibitors, alone or in combination, in advanced hepatocellular carcinoma with macrovascular invasion: a single-centre experience in Taiwan

Background

The presence of vascular invasion is associated with poor survival in advanced hepatocellular carcinoma (HCC). We compared the effectiveness of hepatic arterial infusion chemotherapy (HAIC) and immune checkpoint inhibitors (ICIs), alone or in combination, in patients with advanced HCC.

Methods

We retrospectively reviewed medical records of adult patients with unresectable HCC and macrovascular invasion (MVI) who were treated with HAIC or ICIs alone or in combination at a single centre in Taiwan. Overall tumour response, vascular thrombi response, overall survival (OS) and progression-free survival (PFS) in 130 patients were analysed.

Results

The treatment group showed no significant effect on the overall tumour response [objective response rate (ORR), 22.86% for HAIC, 26.09% for ICI, 50.00% for HAIC+ICI; P=0.111], but showed a significant effect on vessel response (objective response rate of tumour thrombi (ORRT), 38.57% for HAIC, 45.65% for ICI, 78.57% for HAIC+ICI; P=0.023). Post-hoc comparisons followed by Bonferroni correction revealed that vessel ORRT was significantly different between the HAIC+ICI and HAIC groups (P=0.014). A significant effect of treatment group on portal vein tumour thrombus (PVTT) was also detected (ORRT, 40.00% for HAIC, 50.00% for ICI, 90.00% for HAIC; P=0.013), with significant difference between the HAIC+ICI and HAIC groups (P=0.005). Patients treated with HAIC, ICI, and HAIC+ICI respectively had 12-month OS rates of 44.9%, 31.4%, and 67.5% (P=0.127) and 12-month PFS rates of 21.2%, 24.6%, and 33.2% (P=0.091). In multivariate analysis of PFS, HAIC+ICI was associated with reduced risk of progression or death compared with HAIC alone (adjusted hazard ratio: 0.46; 95% confidence interval: 0.23-0.94; P=0.032).

Conclusions

HAIC combined with ICIs had a superior response of PVTT compared to HAIC alone, and was associated with reduced risk of progression or death. Future studies are needed to address the survival benefit of the combination therapy in advanced HCC with MVI.

Association between Biomarkers (VEGF-R2, VEGF-R3, VCAM-1) and Treatment Duration in Patients with Neuroendocrine Tumors Receiving Therapy with First-Generation Somatostatin Analogues

Angiogenic factors (AF) promote vascular formation and may thus support neuroendocrine tumour (NET) development. This study aimed to assess AF serum level changes in NET patients treated with prolonged-acting somatostatin analogues (SSAs). The study enrolled 49 healthy volunteers (Group A) and 56 NET patients: treatment naïve (Group B) and after-SSA treatment in various periods (months): under 12 (Group C), 13–24 (Group D), 25–36 (Group E), 37–60 (Group F), and over 60 months (Group G). The serum vascular endothelial growth factor receptors 2, 3 (VEGF-R2, VEGF-R3), and vascular cell adhesion molecule-1 (VCAM-1) concentrations were tested using the ELISA. We noted significant differences in the concentrations of VEGF-R2, VEGF-R3, and VCAM-1 depending on the SSA treatment duration (p < 0.001). In the studied AFs, the highest decreasing levels of VEGF-R2 were observed after two years of therapy. However, monitoring VEGF-R2, VEGF-R3, and VCAM-1 during SSA treatment did not allow for the identification of good responders for this kind of therapy. Therefore, these biomarker measurements were not helpful in assessing SSA treatment effectiveness in NET patients.

A Single Nucleotide Polymorphism rs1010816 Predicts Sorafenib Therapeutic Outcomes in Advanced Hepatocellular Carcinoma

Sorafenib is currently a targeted agent widely used in the treatment of advanced hepatocellular carcinoma (aHCC). However, to date there is still a lack of a reliable marker capable of predicting sorafenib therapeutic responses. Here, we conducted a genome-wide association study (GWAS) to identify candidate single-nucleotide polymorphism outcome predictors in aHCC patients. A total of 74 real-world sorafenib-treated aHCC patients were enrolled for GWAS and outcome analysis. GWAS showed that rs1010816 (p = 2.2 × 10^-7) was associated with sorafenib therapeutic response in aHCC patients. Kaplan-Meier analysis indicated that the "TT" genotype was significantly associated with a favorable therapeutic response but not significantly associated with overall survival (OS). Univariate followed by multivariate Cox proportional hazard analysis showed that ascites, main portal vein thrombosis, lower platelet count, lower total sorafenib doses, higher PALBI score in model A and higher ALBI grade in model B were significantly associated with a shorter OS. Subgroup analysis showed that only in alcoholic aHCC patients treated by sorafenib, rs1010816 "TT" genotype was significantly associated with longer OS (p = 0.021). Sorafenib had a favorable therapeutic outcome in alcoholic aHCC patients carrying rs1010816 "TT" genotype.

Development and validation of prognostic nomograms for large hepatocellular carcinoma after HAIC

Background and aims

Hepatic arterial infusion chemotherapy (HAIC) using the FOLFOX regimen (oxaliplatin plus fluorouracil and leucovorin) is a promising option for large hepatocellular carcinoma (HCC). However, post-HAIC prognosis can vary in different patients due to tumor heterogeneity. Herein, we established two nomogram models to assess the survival prognosis of patients after HAIC combination therapy.

Methods

A total of 1082 HCC patients who underwent initial HAIC were enrolled between February 2014 and December 2021. We built two nomogram models for survival prediction: the preoperative nomogram (pre-HAICN) using preoperative clinical data and the postoperative nomogram (post-HAICN) based on pre-HAICN and combination therapy. The two nomogram models were internally validated in one hospital and externally validated in four hospitals. A multivariate Cox proportional hazards model was used to identify risk factors for overall survival (OS). The performance outcomes of all models were compared by area under the receiver operating characteristic curve (AUC) analysis with the DeLong test.

Results

Multivariable analysis identified larger tumor size, vascular invasion, metastasis, high albumin-bilirubin grade, and high alpha-fetoprotein as indicators for poor prognosis. With these variables, the pre-HAICN provided three risk strata for OS in the training cohort: low risk (5-year OS, 44.9%), middle risk (5-year OS, 20.6%), and high risk (5-year OS, 4.9%). The discrimination of the three strata was improved significantly in the post-HAICN, which included the above-mentioned factors and number of sessions, combination with immune checkpoint inhibitors, tyrosine kinase inhibitors, and local therapy (AUC, 0.802 versus 0.811, p < 0.001).

Conclusions

The nomogram models are essential to identify patients with large HCC suitable for treatment with HAIC combination therapy and may potentially benefit personalized decision-making.

Lay summary

Hepatic arterial infusion chemotherapy (HAIC) provides sustained higher concentrations of chemotherapy agents in large hepatocellular carcinoma (HCC) by hepatic intra-arterial, result in better objective response outperformed the intravenous administration. HAIC is significantly correlated with favorable survival outcome and obtains extensive support in the effective and safe treatment of intermediate advanced-stage HCC. In view of the high heterogeneity of HCC, there is no consensus regarding the optimal tool for risk stratification before HAIC alone or HAIC combined with tyrosine kinase inhibitors or immune checkpoint inhibitors treatment in HCC. In this large collaboration, we established two nomogram models to estimate the prognosis and evaluate the survival benefits with different HAIC combination therapy. It could help physicians in decision-making before HAIC and comprehensive treatment for large HCC patients in clinical practice and future trials.

Deep Learning-Based Prediction of Molecular Tumor Biomarkers from H&E: A Practical Review

Molecular and genomic properties are critical in selecting cancer treatments to target individual tumors, particularly for immunotherapy. However, the methods to assess such properties are expensive, time-consuming, and often not routinely performed. Applying machine learning to H&E images can provide a more cost-effective screening method. Dozens of studies over the last few years have demonstrated that a variety of molecular biomarkers can be predicted from H&E alone using the advancements of deep learning: molecular alterations, genomic subtypes, protein biomarkers, and even the presence of viruses. This article reviews the diverse applications across cancer types and the methodology to train and validate these models on whole slide images. From bottom-up to pathologist-driven to hybrid approaches, the leading trends include a variety of weakly supervised deep learning-based approaches, as well as mechanisms for training strongly supervised models in select situations. While results of these algorithms look promising, some challenges still persist, including small training sets, rigorous validation, and model explainability. Biomarker prediction models may yield a screening method to determine when to run molecular tests or an alternative when molecular tests are not possible. They also create new opportunities in quantifying intratumoral heterogeneity and predicting patient outcomes.

Deep Learning-Based Classification of Uterine Cervical and Endometrial Cancer Subtypes from Whole-Slide Histopathology Images

Uterine cervical and endometrial cancers have different subtypes with different clinical outcomes. Therefore, cancer subtyping is essential for proper treatment decisions. Furthermore, an endometrial and endocervical origin for an adenocarcinoma should also be distinguished. Although the discrimination can be helped with various immunohistochemical markers, there is no definitive marker. Therefore, we tested the feasibility of deep learning (DL)-based classification for the subtypes of cervical and endometrial cancers and the site of origin of adenocarcinomas from whole slide images (WSIs) of tissue slides. WSIs were split into 360 × 360-pixel image patches at 20× magnification for classification. Then, the average of patch classification results was used for the final classification. The area under the receiver operating characteristic curves (AUROCs) for the cervical and endometrial cancer classifiers were 0.977 and 0.944, respectively. The classifier for the origin of an adenocarcinoma yielded an AUROC of 0.939. These results clearly demonstrated the feasibility of DL-based classifiers for the discrimination of cancers from the cervix and uterus. We expect that the performance of the classifiers will be much enhanced with an accumulation of WSI data. Then, the information from the classifiers can be integrated with other data for more precise discrimination of cervical and endometrial cancers.