Background: Tumor programmed death-ligand 1 (PD-L1) status is useful in determining which patients may benefit from programmed death-1 (PD-1)/PD-L1 inhibitors. However, little is known about the ...association between PD-L1 status and tumor histopathological patterns. Using deep learning, we predicted PD-L1 status from hematoxylin and eosin (H and E) whole-slide images (WSIs) of nonsmall cell lung cancer (NSCLC) tumor samples. Materials and Methods: One hundred and thirty NSCLC patients were randomly assigned to training (n = 48) or test (n = 82) cohorts. A pair of H and E and PD-L1-immunostained WSIs was obtained for each patient. A pathologist annotated PD-L1 positive and negative tumor regions on the training samples using immunostained WSIs for reference. From the H and E WSIs, over 145,000 training tiles were generated and used to train a multi-field-of-view deep learning model with a residual neural network backbone. Results: The trained model accurately predicted tumor PD-L1 status on the held-out test cohort of H and E WSIs, which was balanced for PD-L1 status (area under the receiver operating characteristic curve AUC =0.80, P<< 0.01). The model remained effective over a range of PD-L1 cutoff thresholds (AUC = 0.67-0.81, P ≤ 0.01) and when different proportions of the labels were randomly shuffled to simulate interpathologist disagreement (AUC = 0.63-0.77, P ≤ 0.03). Conclusions: A robust deep learning model was developed to predict tumor PD-L1 status from H and E WSIs in NSCLC. These results suggest that PD-L1 expression is correlated with the morphological features of the tumor microenvironment.
Malaria is a devastating disease that causes significant global morbidity and mortality. The rise of drug resistance against artemisinin-based combination therapy demonstrates the necessity to ...develop alternative antimalarials with novel mechanisms of action. We report the discovery of Ki8751 as an inhibitor of essential kinase PfPK6. 79 derivatives were designed, synthesized and evaluated for PfPK6 inhibition and antiplasmodial activity. Using group efficiency analyses, we established the importance of key groups on the scaffold consistent with a type II inhibitor pharmacophore. We highlight modifications on the tail group that contribute to antiplasmodial activity, cumulating in the discovery of compound 67, a PfPK6 inhibitor (IC50 = 13 nM) active against the P. falciparum blood stage (EC50 = 160 nM), and compound 79, a PfPK6 inhibitor (IC50 < 5 nM) with dual-stage antiplasmodial activity against P. falciparum blood stage (EC50 = 39 nM) and against P. berghei liver stage (EC50 = 220 nM).
Display omitted
•Ki8751, a type II human VEGFR2 inhibitor, was discovered to be a PfPK6 inhibitor.•79 analogues were designed, synthesized, and screened for PfPK6 inhibition.•Top analogues were screened for antiplasmodial activity.•67 inhibits PfPK6 and is active against P. falciparum asexual blood stage.•79 inhibits PfPK6 and is active against Plasmodium liver and asexual blood stages.
Essential plasmodial kinases PfGSK3 and PfPK6 are considered novel drug targets to combat rising resistance to traditional antimalarial therapy. Herein, we report the discovery of IKK16 as a dual ...PfGSK3/PfPK6 inhibitor active against blood stage Pf3D7 parasites. To establish structure–activity relationships for PfPK6 and PfGSK3, 52 analogues were synthesized and assessed for the inhibition of PfGSK3 and PfPK6, with potent inhibitors further assessed for activity against blood and liver stage parasites. This culminated in the discovery of dual PfGSK3/PfPK6 inhibitors 23d (PfGSK3/PfPK6 IC50 = 172/11 nM) and 23e (PfGSK3/PfPK6 IC50 = 97/8 nM) with antiplasmodial activity (23d Pf3D7 EC50 = 552 ± 37 nM and 23e Pf3D7 EC50 = 1400 ± 13 nM). However, both compounds exhibited significant promiscuity when tested in a panel of human kinase targets. Our results demonstrate that dual PfPK6/PfGSK3 inhibitors with antiplasmodial activity can be identified and can set the stage for further optimization efforts.
To achieve minimum DNA input requirements for next-generation sequencing (NGS), pathologists visually estimate macrodissection and slide count decisions. Unfortunately, misestimation may cause tissue ...waste and increased laboratory costs. We developed an artificial intelligence (AI)-augmented smart pathology review system (SmartPath) to empower pathologists with quantitative metrics for accurately determining tissue extraction parameters. SmartPath uses two deep learning architectures, a U-Net based network for cell segmentation and a multi-field-of-view convolutional network for tumor area segmentation, to extract features from digitized H&E-stained formalin-fixed paraffin-embedded slides. From the segmented tumor area, SmartPath suggests a macrodissection area. To predict DNA yield per slide, the extracted features from within the macrodissection area are correlated with known DNA yields to fit a regularized linear model (R = 0.85). Then, a pathologist-defined target yield divided by the predicted DNA yield per slide gives the number of slides to scrape. Following model development, an internal validation trial was conducted within the Tempus Labs molecular sequencing laboratory. We evaluated our system on 501 clinical colorectal cancer slides, where half received SmartPath-augmented review and half traditional pathologist review. The SmartPath cohort had 25% more DNA yields within a desired target range of 100-2000 ng. The number of extraction attempts was statistically unchanged between cohorts. The SmartPath system recommended fewer slides to scrape for large tissue sections, saving tissue in these cases. Conversely, SmartPath recommended more slides to scrape for samples with scant tissue sections, especially those with degraded DNA, helping prevent costly re-extraction due to insufficient extraction yield. A statistical analysis was performed to measure the impact of covariates on the results, offering insights on how to improve future applications of SmartPath. With these improvements, AI-augmented histopathologic review has the potential to decrease tissue waste, sequencing time, and laboratory costs by optimizing DNA yields, especially for samples with scant tissue and/or degraded DNA.
Clinical decision-making in oncology involves multimodal data such as radiology scans, molecular profiling, histopathology slides, and clinical factors. Despite the importance of these modalities ...individually, no deep learning framework to date has combined them all to predict patient prognosis. Here, we predict the overall survival (OS) of glioma patients from diverse multimodal data with a Deep Orthogonal Fusion (DOF) model. The model learns to combine information from multiparametric MRI exams, biopsy-based modalities (such as H&E slide images and/or DNA sequencing), and clinical variables into a comprehensive multimodal risk score. Prognostic embeddings from each modality are learned and combined via attention-gated tensor fusion. To maximize the information gleaned from each modality, we introduce a multimodal orthogonalization (MMO) loss term that increases model performance by incentivizing constituent embeddings to be more complementary. DOF predicts OS in glioma patients with a median C-index of 0.788 +/- 0.067, significantly outperforming (p=0.023) the best performing unimodal model with a median C-index of 0.718 +/- 0.064. The prognostic model significantly stratifies glioma patients by OS within clinical subsets, adding further granularity to prognostic clinical grading and molecular subtyping.
To achieve minimum DNA input and tumor purity requirements for next-generation sequencing (NGS), pathologists visually estimate macrodissection and slide count decisions. Misestimation may cause ...tissue waste and increased laboratory costs. We developed an AI-augmented smart pathology review system (SmartPath) to empower pathologists with quantitative metrics for determining tissue extraction parameters. Using digitized H&E-stained FFPE slides as inputs, SmartPath segments tumors, extracts cell-based features, and suggests macrodissection areas. To predict DNA yield per slide, the extracted features are correlated with known DNA yields. Then, a pathologist-defined target yield divided by the predicted DNA yield/slide gives the number of slides to scrape. Following model development, an internal validation trial was conducted within the Tempus Labs molecular sequencing laboratory. We evaluated our system on 501 clinical colorectal cancer slides, where half received SmartPath-augmented review and half traditional pathologist review. The SmartPath cohort had 25% more DNA yields within a desired target range of 100-2000ng. The SmartPath system recommended fewer slides to scrape for large tissue sections, saving tissue in these cases. Conversely, SmartPath recommended more slides to scrape for samples with scant tissue sections, helping prevent costly re-extraction due to insufficient extraction yield. A statistical analysis was performed to measure the impact of covariates on the results, offering insights on how to improve future applications of SmartPath. Overall, the study demonstrated that AI-augmented histopathologic review using SmartPath could decrease tissue waste, sequencing time, and laboratory costs by optimizing DNA yields and tumor purity.
Background: Obesity is rapidly increasing with disparately high rates in the Black/African-American population. However, there are limited body composition estimation methods that are accessible and ...costeffective for those in underserved populations or remote geographical locations. Thus, mobile body composition estimations that employ the remote healthcare model are urgently needed. Methods: The purpose of this study was to determine the agreement between smartphone-based visual body composition (VBC) applications and a 4-compartment (4C) model in Black/African-American individuals. Thirty-six (F=24, M=12) participants completed several body composition assessments including those necessary for 4C model estimation and automated assessments produced from two smartphone applications (Amazon Halo and myBVI) using either Apple or Samsung phones. Body fat % (%fat), fat-free mass (FFM), and fat mass (FM) were compared to a 4C model. Results: There were no significant mean differences between the 4C model and HaloApple, HaloSamsung, or myBVI for any variable (p>0.05). All body composition variables were very strongly correlated with 4C model estimations (r>0.90, p<0.001) except for %fat from myBVI (r=0.70, p<0.001). For %fat, HaloApple and HaloSamsung had a total error (TE) of 3.9%, which was lower than for myBVI (TE=6.4%). myBVI showed significant proportional bias for %fat (ß=0.34, p=0.023) which was not observed for Halo (p>0.05). For FFM, HaloApple and HaloSamsung had a TE of 2.9 kg and 3.0 kg, respectively, and TE for myBVI was 6.4 kg. Proportional biases for estimates of FFM were observed for HaloApple (ß=0.089, p=0.002) and HaloSamsung (ß=0.097, p=0.001) which was not observed for myBVI (ß=0.043, p=0.531). Conclusions: VBC estimates using Amazon Halo show fairly good to good agreement with a 4C model for %fat and FFM in Black/AfricanAmerican individuals but may overestimate FFM relative to 4C in this group. myBVI demonstrated lower validity for body composition estimation in Black/African-American males and females.
A generalized computational methodology for reduced order acoustic‐structural coupled modeling of the aeroacoustics of a wind turbine blade is presented. This methodology is used to investigate the ...acoustic pressure distribution in and around airfoils to guide the development of a passive damage detection approach for structural health monitoring of wind turbine blades for the first time. The output of a k − ε turbulence model computational fluid dynamics simulation is used to calculate simple acoustic sources on the basis of model tuning with published experimental data. The methodology is then applied to a computational case study of a 0.3048‐m chord NACA 0012 airfoil with two internal cavities, each with a microphone placed along the shear web. Five damage locations and four damage sizes are studied and compared with the healthy baseline case for three strategically selected acoustic frequencies: 1, 5, and 10 kHz. In 22 of the 36 cases in which the front cavity is damaged, the front cavity microphone measures an increase in sound pressure level (SPL) above 3 dB, while rear cavity damage only results in six out of 24 cases with a 3‐dB increase in the rear cavity. The 1‐ and 5‐kHz cases show a more consistent increase in SPL than the 10‐kHz case, illustrating the spectral dependency of the model. The case study shows how passive acoustic detection could be used to identify blade damage, while providing a template for application of the methodology to investigate the feasibility of passive detection for any specific turbine blade.