Abstract Tumor infiltrating lymphocytes (TILs)-based biomarkers have emerged as a robust method for predicting treatment efficacy and outcome in colorectal cancer (CRC). However, the quantification of TILs remains a strenuous task for pathologists and is susceptible to inter-pathologist variations. Deep learning (DL) can predict cancer biomarkers directly from routine hematoxylin and eosin (H&E) pathology slides, enabling the automated and consistent quantification of TILs for clinical decision-making. Our study focuses on the prediction of TILs per high-power field (HPF) on two large cohorts of CRC patients. We developed a weakly-supervised, transformer-based deep regression model to predict TILs per HPF from routine H&E-stained histology slides. We trained the model using 5-fold cross-validation on a large cohort of CRC patients (n=1,738), and validated its performance on an external cohort of CRC patients (n=223). The model was evaluated using the Pearson’s correlation coefficient r. Moreover, pathologist-determined TILs per HPF was categorized into low (<2) and high (≥2), enabling evaluation via the area under the receiver operating characteristic curve (AUROC). Our model achieved a significant correlation coefficient of 0.60 (p<0.0001) for the predicted TILs per HPF in the holdout test sets of the internal cohort, and a significant correlation coefficient of 0.57 (p<0.0001) on the external validation cohort. Using the thresholds for low and high TILs per HPF, we found that the model accurately predicted the presence of high TILs per HPF in both the internal and external test cohorts, yielding AUROCs of 0.76 and 0.81, respectively. These findings underscore the efficacy of our deep learning model in predicting TILs per HPF from routine pathology slides. This approach holds promise for cost-effective, efficient and uniform quantification of TILs per HPF in CRC, potentially across various cancer types. Citation Format: Omar S. El Nahhas, Joseph D. Bonner, Joel K. Greenson, Daniel B. Schmolze, Lawrence Shaktah, Jonathan Salazar, Lorena Reynaga, Sidney Lindsey, Jenny Lu, Victor Moreno, Stephanie L. Schmit, Ya-Yu Tsai, Stanley R. Hamilton, Gad Rennert, Jakob N. Kather, Stephen B. Gruber. Weakly-supervised prediction of tumor infiltrating lymphocytes per high power field from colorectal cancer histopathology slides using regression transformers [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(7_Suppl):Abstract nr LB386.

Abstract Tumor infiltrating lymphocytes (TILs) per high-power field (HPF) as measured by expert pathologists is a useful, independent, validated prognostic marker in colorectal cancer (CRC). Artificial intelligence (AI) techniques of deep learning (DL) can predict TILs directly from routinely collected, digitized hematoxylin and eosin (H&E) pathology slides, but AI-predicted TILs have not yet been thoroughly validated as clinically relevant biomarker. Here we test whether AI-predicted TILs per HPF can be validated as an independent prognostic biomarker of CRC-specific survival in patients. We constructed an AI model to predict TILS per HPF using a weakly-supervised, transformer-based deep regression model, trained using 5-fold cross validation in a large dataset (n=1,738) and validated in a second independent dataset (n=223) of CRC cases. The model predicted TILs per HPF as a continuous variable, which was carried forward for survival analyses both as a continuous variable and dichotomized at a previously established a priori threshold of <2 vs. ≥2 TILs per HPF. Using longitudinal data from the population-based Molecular Epidemiology of Colorectal Cancer Study (median follow-up = 95 months), survival analyses were performed using non-parametric and Cox-proportional hazards models, with and without adjustment for age, sex, ethnicity, stage, and molecularly measured microsatellite instability (MSI). Two or more AI-predicted TILs per HPF were significantly associated with 5-year CRC-specific survival (p=0.0000037), 5-year overall survival (p=0.00021), and overall survival (p=0.000067). In a Cox proportional hazards model adjusting for age, sex, ethnicity, stage, and MSI, ≥2 AI-predicted TILs per HPF was significantly associated with improved 5-year CRC-specific survival, with a hazard ratio (HR) = 0.62, (95% confidence interval; 0.43, 0.91), p=0.01. Our new AI-driven deep learning model, which we call HopeSTIL, provides a highly efficient algorithm for analyzing digital pathology images of H&E sections of CRC to quantify TILs per HPF. HopeSTIL is a validated prognostic marker for 5-year CRC-specific survival that does not require a pathologist to manually count and score TILs per HPF. Citation Format: Stephen B. Gruber, Omar S. El Nahhas, Joseph D. Bonner, Joel K. Greenson, Daniel Schmolze, Lawrence Shaktah, Jonathan Salazar, Lorena Reynaga, Sidney Lindsey, Jenny Lu, Allen Mao, Victor Moreno, Stephani L. Schmit, Ya-Yu Tsai, Stanley R. Hamilton, Gad Rennert, Jakob N. Kather. Artificial intelligence measures of tumor infiltrating lymphocytes predict colorectal cancer-specific and overall survival [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(7_Suppl):Abstract nr LB384.

Supplementary Table 1, Figures 1-3, Methods from Microenvironmental Independence Associated with Tumor Progression

Mono-hydroxychlorins are uncommon macrocycles that have only been synthetically realized by modifying porphyrin rings using the harsh oxidizing agent OsO4. We show here that a more directed delivery of the mono-hydroxychlorin may be concomitantly obtained from the oxidation of porphyrinogen using the mild conditions of the high dilution Lindsey porphyrin forming reaction where water content is minimized by using dry CHCl3 within the environment of a glovebox. We now report the direct synthesis of 17,18-dihydro-18-hydroxy-5,10,15,20-tetrakis-(4-fluoro,2,6-dimethylphenyl)-porphyrin (2H-TFChl-OH) together with the corresponding freebase porphyrin TFP. The TFP has been metalated with FeBr2 and MgBr2 center dot OEt2 resulting in metalloporphyrins Fe(III)TFP(Cl) and Mg(II)-TFP which have been structurally characterized by single-crystal X-ray crystallography. We find that the excited state properties of the mono-hydroxychlorin are similar to that of its parent TFP and Mg(II)TFP porphyrin congeners. Excited state deactivation by vibronic coupling to the high energy O-H oscillator is circumvented with the hydroxyl group remote to the 18-electron framework of the chlorin ring. These results reveal that strong H-bonding groups may be introduced on the periphery of the chlorin ring while maintaining the light-gathering properties that lie at the heart of photosynthesis of the chlorin ring. 17,18-dihydro-18-hydroxy-5,10,15,20-tetrakis-(4-fluoro,2,6-dimethylphenyl)-porphyrin together with the corresponding freebase porphyrin were obtained under mild conditions. The freebase porphyrin (TFP) was metalated with FeBr2 and MgBr2 center dot OEt2, resulting in the metalloporphyrins Fe(III)TFP(Cl) and Mg(II)TFP which have been structurally characterized by single-crystal X-ray crystallography. We found that the excited state properties of the mono-hydroxychlorin are similar to that of its parent TFP and Mg(II)TFP porphyrin congeners.

Free Article from Microenvironmental Independence Associated with Tumor Progression

10597 Background: Lynch syndrome (LS) is the most common cause of hereditary colorectal cancer (CRC) with an increased CRC lifetime risk of 70-80%. LS affects 1:250 individuals and is caused by pathogenic variants in the mismatch repair (MMR) genes. Statistical prediction models such as MMRpro and PREMM5 are widely used to identify LS carriers. However, these models are trained and validated in mostly white populations, and there remains a gap in understanding their performance in Hispanic populations. The purpose of this study was to evaluate the performance of MMRpro and PREMM5 on a large Hispanic cohort from the Clinical Cancer Genomics Community Research Network (CCGCRN). Methods: We validated MMRpro and PREMM5 on 3,490 CCGCRN families, of which 1,122 are Hispanic and 2,062 Non-Hispanic. The two models were evaluated for discrimination using the C-statistic, calibration using the observed to expected ratio (O/E), and overall performance using the root Brier score and negative and positive predictive value (NPV/PPV) at the 5% carrier probability threshold. Evaluations were stratified by ethnicity, and 95% confidence intervals were obtained via bootstrapping for all measures. Results: The C-statistic is 0.90 for both MMRpro (95% CI: 0.88, 0.92) and PREMM5 (95% CI: 0.87, 0.92). When stratified by ethnicity, the C-statistics are 0.96 (95% CI: 0.94, 0.97) and 0.86 (95% CI: 0.83, 0.89) for Hispanics and Non-Hispanics, respectively, in MMRpro, and 0.96 (95% CI: 0.94, 0.97) and 0.84 (95% CI: 0.79, 0.88) in PREMM5. Both models underpredict mutation probabilities, with O/E ratios ranging from 1.79 to 1.96. At a 5% threshold, variations in PPV between Hispanics and Non-Hispanics are observed in both models: 0.72 (95% CI: 0.63, 0.80) and 0.43 (95% CI: 0.37, 0.50) in Hispanic and Non-Hispanic groups in MMRpro; 0.50 (95% CI: 0.43, 0.57) and 0.25 (95% CI: 0.20, 0.30) in PREMM5. We observe less variation and higher values in NPVs in both models. Conclusions: Overall, MMRpro and PREMM5 perform well in this cohort in predicting the probability of having a pathogenic variant in an MMR gene, with modest underprediction. While these results offer reassurance for the clinical use of MMRpro and PREMM5 in Hispanic populations, further validation studies in underrepresented racial and ethnic populations are crucial.

We have developed a novel, portable, gravity-fed, microfluidics-based platform suitable for optical interrogation of long-term organotypic cell culture. This system is designed to provide convenient control of cell maintenance, nutrients, and experimental reagent delivery to tissue-like cell densities housed in a transparent, low-volume microenvironment. To demonstrate the ability of our Thick-Tissue Bioreactor (TTB) to provide stable, long-term maintenance of high-density cellular arrays, we observed the morphogenic growth of human mammary epithelial cell lines, MCF-10A and their invasive variants, cultured under three-dimensional (3D) conditions inside our system. Over the course of 21 days, these cells typically develop into hollow "mammospheres" if cultured in standard 3D Matrigel. This complex morphogenic process requires alterations in a variety of cellular functions, including degradation of extracellular matrix that is regulated by cell-produced matrix proteinases. For our "drug" delivery testing and validation experiments we have introduced proteinase inhibitors into the fluid supply system, and we observed both reduced proteinase activity and inhibited cellular morphogenesis. The size inhibition results correlated well with the overall proteinase activities of the tested cells.

Cell migration paths of mammary epithelial cells (expressing different versions of the promigratory tyrosine kinase receptor Her2/Neu) were analyzed within a bimodal framework that is a generalization of the run-and-tumble description applicable to bacterial migration. The mammalian cell trajectories were segregated into two types of alternating modes, namely, the "directional mode'' (mode I, the more persistent mode, analogous to the bacterial run phase) and the "re-orientation mode'' (mode II, the less persistent mode, analogous to the bacterial tumble phase). Higher resolution (more pixel information, relative to cell size) and smaller sampling intervals (time between images) were found to give a better estimate of the deduced single cell dynamics (such as directional-mode time and turn angle distribution) of the various cell types from the bimodal analysis. The bimodal analysis tool permits the deduction of short-time dynamics of cell motion such as the turn angle distributions and turn frequencies during the course of cell migration compared to standard methods of cell migration analysis. We find that the 2-h mammalian cell tracking data do not fall into the diffusive regime implying that the often-used random motility expressions for mammalian cell motion (based on assuming diffusive motion) are invalid over the time steps (fraction of minute) typically used in modeling mammalian cell migration.

Objective: Our objective was to compare a newly developed semiquantitative visual scoring (SVS) method with the current standard, the Response Evaluation Criteria in Solid Tumors (RECIST) method, in the categorization of treatment response and reader agreement for patients with metastatic lung cancer followed by computed tomography. Materials and Methods: The 18 subjects (5 women and 13 men; mean age, 62.8 years) were from an institutional review board-approved phase 2 study that evaluated a second-line chemotherapy regimen for metastatic (stages III and IV) non-small cell lung cancer. Four radiologists, blinded to the patient outcome and each other's reads, evaluated the change in the patients' tumor burden from the baseline to the first restaging computed tomographic scan using either the RECIST or the SVS method. We compared the numbers of patients placed into the partial response, the stable disease (SD), and the progressive disease (PD) categories (Fisher exact test) and observer agreement (&kgr; statistic). Results: Requiring the concordance of 3 of the 4 readers resulted in the RECIST placing 17 (100%) of 17 patients in the SD category compared with the SVS placing 9 (60%) of 15 patients in the partial response, 5 (33%) of the 15 patients in the SD, and 1 (6.7%) of the 15 patients in the PD categories (P < 0.0001). Interobserver agreement was higher among the readers using the SVS method (&kgr;, 0.54; P < 0.0001) compared with that of the readers using the RECIST method (&kgr;, −0.01; P = 0.5378). Conclusions: Using the SVS method, the readers more finely discriminated between the patient response categories with superior agreement compared with the RECIST method, which could potentially result in large differences in early treatment decisions for advanced lung cancer.