This is a repository copy of Pan-cancer image-based detection of clinically actionable genetic alterations.

and tluedde@ukaachen.de 34 35 Precision treatment of cancer relies on genetic alterations which are diagnosed by molecular 36 biology assays. 1 These tests can be a bottleneck in oncology workflows because of high turna-37 round time, tissue usage and costs. 2 Here, we show that deep learning can predict point muta-38 tions, molecular tumor subtypes and immune-related gene expression signatures 3,4 directly 39 from routine histological images of tumor tissue. We developed and systematically optimized 40 a one-stop-shop workflow and applied it to more than 4000 patients with breast 5 , colon and 41 rectal 6 , head and neck 7 , lung 8,9 , pancreatic 10 , prostate 11 cancer, melanoma 12 and gastric 13 can-42 cer. Together, our findings show that a single deep learning algorithm can predict clinically ac-43 tionable alterations from routine histology data. Our method can be implemented on mobile 44 hardware 14 , potentially enabling point-of-care diagnostics for personalized cancer treatment 45 in individual patients. 46 Clinical guidelines recommend molecular testing of tumor tissue for most patients with advanced 47 209 The results are in part based upon data generated by the TCGA Research Network: http://can-210 cergenome.nih.gov/. Our funding sources are as follows. J.N.K.: RWTH University Aachen (START 211

The results of this study demonstrate that muscle function progressively declines in the Cy/+ rat model of CKD. Because whole muscle mass and architecture do not vary between CKD and NL, but CKD muscles show reduction in individual fiber CSA, our data suggest that the functional decline is related to increased muscle fiber atrophy.

A model’s ability to express its own predictive uncertainty is an essential attribute for maintaining clinical user confidence as computational biomarkers are deployed into real-world medical settings. In the domain of cancer digital histopathology, we describe a clinically-oriented approach to uncertainty quantification for whole-slide images, estimating uncertainty using dropout and calculating thresholds on training data to establish cutoffs for low- and high-confidence predictions. We train models to identify lung adenocarcinoma vs. squamous cell carcinoma and show that high-confidence predictions outperform predictions without uncertainty, in both cross-validation and testing on two large external datasets spanning multiple institutions. Our testing strategy closely approximates real-world application, with predictions generated on unsupervised, unannotated slides using predetermined thresholds. Furthermore, we show that uncertainty thresholding remains reliable in the setting of domain shift, with accurate high-confidence predictions of adenocarcinoma vs. squamous cell carcinoma for out-of-distribution, non-lung cancer cohorts.

Sickle cell disease (SCD) patients are at a four- to 100-fold increased risk for thrombosis compared with the general population, although the mechanisms and risk factors are not clear. We investigated the incidence and predictors for thrombosis in a retrospective, longitudinal cohort of 1193 pediatric and adult SCD patients treated at our institution between January 2008 and December 2017. SCD diagnosis and thrombotic complications were identified using International Classification of Diseases coding and verified through medical chart review. Clinical and laboratory data were extracted from the medical records. With a median follow-up of 6.4 years, 208 (17.4%) SCD patients experienced 352 thrombotic events (64 strokes, 288 venous thromboembolisms [VTE]). Risk factors for stroke included older age and HbSS/Sβ0-genotype and a lower hemoglobin (Hb) F% in the subset of HbSS/Sβ0-genotype patients (P < .05). VTE risk was independently associated with lower estimated glomerular filtration rate, hydroxyurea (HU) use, HbSS/Sβ0 genotype, and higher white blood cell (WBC) counts and Hb (P ≤ .03). Two thrombomodulin gene variants previously associated with thrombosis in the general African American population, THBD rs2567617 (minor allele frequency [MAF] 0.25; odds ratio [OR], 1.5; P = .049) and THBD rs1998081 (MAF, 0.24; OR, 1.5; P = .059), were associated with thrombosis in this cohort. In summary, thrombotic complications are common, and several traditional and SCD-specific risk factors are associated with thrombotic risk. Future studies integrating clinical, laboratory, and genetic risk factors may improve our understanding of thrombosis and guide intervention practices in SCD.

1Oncogenic viruses like human papilloma virus (HPV) or Epstein Barr virus (EBV) are a major cause of human 2 cancer. Viral oncogenesis has a direct impact on treatment decisions because virus-associated tumors can 3 demand a lower intensity of chemotherapy and radiation or can be more susceptible to immune check-4 point inhibition. However, molecular tests for HPV and EBV are not ubiquitously available. 5We hypothesized that the histopathological features of virus-driven and non-virus driven cancers are suf-6 ficiently different to be detectable by artificial intelligence (AI) through deep learning-based analysis of 7 images from routine hematoxylin and eosin (HE) stained slides. We show that deep transfer learning can 8 predict presence of HPV in head and neck cancer with a patient-level 3-fold cross validated area-under-9 the-curve (AUC) of 0.89 [0.82; 0.94]. The same workflow was used for Epstein-Barr virus (EBV) driven 10 gastric cancer achieving a cross-validated AUC of 0.80 [0.70; 0.92] and a similar performance in external 11 validation sets. Reverse-engineering our deep neural networks, we show that the key morphological fea-12 tures can be made understandable to humans. 13 This workflow could enable a fast and low-cost method to identify virus-induced cancer in clinical trials or 14 clinical routine. At the same time, our approach for feature visualization allows pathologists to look into 15 the black box of deep learning, enabling them to check the plausibility of computer-based image classifi-16 cation. 17 18

Artificial intelligence methods including deep neural networks (DNN) can provide rapid molecular classification of tumors from routine histology with accuracy that matches or exceeds human pathologists. Discerning how neural networks make their predictions remains a significant challenge, but explainability tools help provide insights into what models have learned when corresponding histologic features are poorly defined. Here, we present a method for improving explainability of DNN models using synthetic histology generated by a conditional generative adversarial network (cGAN). We show that cGANs generate high-quality synthetic histology images that can be leveraged for explaining DNN models trained to classify molecularly-subtyped tumors, exposing histologic features associated with molecular state. Fine-tuning synthetic histology through class and layer blending illustrates nuanced morphologic differences between tumor subtypes. Finally, we demonstrate the use of synthetic histology for augmenting pathologist-in-training education, showing that these intuitive visualizations can reinforce and improve understanding of histologic manifestations of tumor biology.

Reference point indentation (RPI) was developed to measure material-level mechanical properties of bone in vivo. Studies using RPI in vivo have discriminated between human subjects with previous skeletal fractures and those without and among dogs given different anti-remodeling drugs. Recently, this technology was extended to rats, providing the first in vivo data for rodents. The goal of the present study was to perform in vivo RPI measurements in mice, the most common animal model used to study bone. Twelve 16-week-old female C57BL/6 mice were subjected to RPI (three tests) on the anterior tibia, followed by a repeat test session on the contralateral limb 28 days later. A custom MATLAB program was used to derive several outcome parameters from RPI force-displacement curves: first cycle indentation distance (ID-1st), ID increase (IDI), total ID (TID), first cycle unloading slope (US-1st) and first cycle energy dissipation (ED-1st). Data within an individual were averaged across the three tests for each time point. Within-animal variation of all RPI parameters on day 1 ranged from 12.8 to 33.4% and from 14.1 to 22.4% on day 28. Between-animal variation on day 1 ranged from 11.4% to 22.8% and from 7.5% to 24.7% on day 28. At both time points, within-and between-animals, US-1st was the least variable parameter and IDI was most variable. All parameters were nonsignificantly lower at day 28 compared with day 1. These data are important to demonstrate the feasibility of collecting bone material property data longitudinally in mice and will inform the design of future studies in terms of statistical power and appropriate sample size considerations.