Building a trustworthy AI differential diagnosis application for Crohn's disease and intestinal tuberculosis.

BACKGROUND: Differentiating between Crohn's disease (CD) and intestinal tuberculosis (ITB) with endoscopy is challenging. We aim to perform more accurate endoscopic diagnosis between CD and ITB by building a trustworthy AI differential diagnosis application. METHODS: A total of 1271 electronic health record (EHR) patients who had undergone colonoscopies at Peking Union Medical College Hospital (PUMCH) and were clinically diagnosed with CD (n = 875) or ITB (n = 396) were used in this study. We build a workflow to make diagnoses with EHRs and mine differential diagnosis features; this involves finetuning the pretrained language models, distilling them into a light and efficient TextCNN model, interpreting the neural network and selecting differential attribution features, and then adopting manual feature checking and carrying out debias training. RESULTS: The accuracy of debiased TextCNN on differential diagnosis between CD and ITB is 0.83 (CR F1: 0.87, ITB F1: 0.77), which is the best among the baselines. On the noisy validation set, its accuracy was 0.70 (CR F1: 0.87, ITB: 0.69), which was significantly higher than that of models without debias. We also find that the debiased model more easily mines the diagnostically significant features. The debiased TextCNN unearthed 39 diagnostic features in the form of phrases, 17 of which were key diagnostic features recognized by the guidelines. CONCLUSION: We build a trustworthy AI differential diagnosis application for differentiating between CD and ITB focusing on accuracy, interpretability and robustness. The classifiers perform well, and the features which had statistical significance were in agreement with clinical guidelines.

Identification and characterization of two DMD pedigrees with large inversion mutations based on a long-read sequencing pipeline.

Pathogenic large inversions are rarely reported on DMD gene due to the lack of effective detection methods. Here we report two DMD pedigrees and proposed a reliable pipeline to define large inversions in DMD patients. In the first pedigree, conventional approaches including multiplex ligation-dependent probe amplification, and whole-exome sequencing by next generation sequencing were failed to detect any pathologic variant. Then an advanced analysis pipeline which consists of RNA-seq, cDNA array capture sequencing, optical mapping, long-read sequencing was built. RNA-seq and cDNA capture sequencing showed a complete absence of transcripts of exons 3-55. Optical mapping identified a 55 Mb pericentric inversion between Xp21 and Xq21. Subsequently, long-read sequencing and Sanger sequencing determined the inversion breakpoints at 32,915,769 and 87,989,324 of the X chromosomes. In the second pedigree, long-read sequencing was directly conducted and Sanger sequencing was performed to verify the mutation. Long-read sequencing and Sanger sequencing found breakpoints at 32,581,576 and 127,797,236 on DMD gene directly. In conclusion, large inversion might be a rare but important mutation type in DMD gene. An effective pipeline was built in detecting large inversion mutations based on long-read sequencing platforms.

Effects of oncolytic viruses and viral vectors on immunity in glioblastoma.

Glioblastoma (GBM) is regarded as an incurable disease due to its poor prognosis and limited treatment options. Virotherapies were once utilized on cancers for their oncolytic effects. And they are being revived on GBM treatment, as accumulating evidence presents the immunogenic effects of virotherapies in remodeling immunosuppressive GBM microenvironment. In this review, we focus on the immune responses induced by oncolytic virotherapies and viral vectors in GBM. The current developments of GBM virotherapies are briefly summarized, followed by a detailed depiction of their immune response. Limitations and lessons inferred from earlier experiments and trials are discussed. Moreover, we highlight the importance of engaging the immune responses induced by virotherapies into the multidisciplinary management of GBM.

Corrigendum: Sequence and Structure Characteristics of 22 Deletion Breakpoints in Intron 44 of the DMD Gene Based on Long-Read Sequencing.

[This corrects the article DOI: 10.3389/fgene.2021.638220.].

Sequence and Structure Characteristics of 22 Deletion Breakpoints in Intron 44 of the DMD Gene Based on Long-Read Sequencing.

Purpose: Exon deletions make up to 80% of mutations in the DMD gene, which cause Duchenne and Becker muscular dystrophy. Exon 45-55 regions were reported as deletion hotspots and intron 44 harbored more than 25% of deletion start points. We aimed to investigate the fine structures of breakpoints in intron 44 to find potential mechanisms of large deletions in intron 44. Methods: Twenty-two dystrophinopathy patients whose deletion started in intron 44 were sequenced using long-read sequencing of a DMD gene capture panel. Sequence homology, palindromic sequences, and polypyrimidine sequences were searched at the breakpoint junctions. RepeatMasker was used to analyze repetitive elements and Mfold was applied to predict secondary DNA structure. Results: With a designed DMD capture panel, 22 samples achieved 2.25 gigabases and 1.28 million reads on average. Average depth was 308× and 99.98% bases were covered at least 1×. The deletion breakpoints in intron 44 were scattered and no breakpoints clustered in any region less than 500 bp. A total of 72.7% of breakpoints located in distal 100 kb of intron 44 and more repetitive elements were found in this region. Microhomologies of 0-1 bp were found in 36.4% (8/22) of patients, which corresponded with non-homologous end-joining. Microhomologies of 2-20 bp were found in 59.1% (13/22) of patients, which corresponded with microhomology-mediated end-joining. Moreover, a 7 bp insertion was found in one patient, which might be evidence of aberrant replication origin firing. Palindromic sequences, polypyrimidine sequences, and small hairpin loops were found near several breakpoint junctions. No evidence of large hairpin loop formation in deletion root sequences was observed. Conclusion: This study was the first to explore possible mechanisms underlying exon deletions starting from intron 44 of the DMD gene based on long-read sequencing. Diverse mechanisms might be associated with deletions in the DMD gene.

A Comprehensive Analysis of 2013 Dystrophinopathies in China: A Report From National Rare Disease Center.

Duchenne muscular dystrophy (DMD) and Becker muscular dystrophy (BMD) are X-linked recessive neuromuscular disorders caused by mutations in DMD. A high-quality database of DMD/BMD is essential not only for clinical practice but also for fundamental research. Here, we aimed to build the largest Chinese national dystrophinopathy database using the National Rare Diseases Registry System of China. Peking Union Medical College Hospital (PUMCH) was the National Rare Diseases Center of China. This research involved 2013 patients with dystrophinopathies, whose diagnoses were confirmed; they were registered and followed up at PUMCH from March 2011 to December 2018. Family history, clinical signs, and treatment data were reported for patients with DMD and BMD at different rates. All six serum biochemical indexes could accurately distinguish between DMD and BMD patients. Copy number variations were the most frequent mutation type (79.2% in DMD and 84.3% in BMD), of which large deletions accounted for 88.4 and 88.6%, large duplications accounted for 11.6 and 11.4% in DMD and BMD, respectively. An exon deletion hotspot, located in exons 45-54, was observed in DMD, and intron 44 was the most frequent deletion starting point (26.5%). Duplication and single nucleotide variations appeared to be uniformly distributed among all exons. Eleven patients were identified to have ultrarare mutation types. Eleven other patients suffered from two separate mutations simultaneously, some of which may have taken place via dependent mechanisms. Thus, we have established the largest hospital-based Chinese dystrophinopathy database via the National Rare Diseases Registry System. This study provides valuable information for further diagnostic and therapeutic studies of dystrophinopathy.

Can natural language processing help differentiate inflammatory intestinal diseases in China? Models applying random forest and convolutional neural network approaches.

BACKGROUND: Differentiating between ulcerative colitis (UC), Crohn's disease (CD) and intestinal tuberculosis (ITB) using endoscopy is challenging. We aimed to realize automatic differential diagnosis among these diseases through machine learning algorithms. METHODS: A total of 6399 consecutive patients (5128 UC, 875 CD and 396 ITB) who had undergone colonoscopy examinations in the Peking Union Medical College Hospital from January 2008 to November 2018 were enrolled. The input was the description of the endoscopic image in the form of free text. Word segmentation and key word filtering were conducted as data preprocessing. Random forest (RF) and convolutional neural network (CNN) approaches were applied to different disease entities. Three two-class classifiers (UC and CD, UC and ITB, and CD and ITB) and a three-class classifier (UC, CD and ITB) were built. RESULTS: The classifiers built in this research performed well, and the CNN had better performance in general. The RF sensitivities/specificities of UC-CD, UC-ITB, and CD-ITB were 0.89/0.84, 0.83/0.82, and 0.72/0.77, respectively, while the values for the CNN of CD-ITB were 0.90/0.77. The precisions/recalls of UC-CD-ITB when employing RF were 0.97/0.97, 0.65/0.53, and 0.68/0.76, respectively, and when employing the CNN were 0.99/0.97, 0.87/0.83, and 0.52/0.81, respectively. CONCLUSIONS: Classifiers built by RF and CNN approaches had excellent performance when classifying UC with CD or ITB. For the differentiation of CD and ITB, high specificity and sensitivity were achieved as well. Artificial intelligence through machine learning is very promising in helping unexperienced endoscopists differentiate inflammatory intestinal diseases. CONFERENCE: The abstract of this article has won the first prize of the Young Investigator Award during the Asian Pacific Digestive Week (APDW) 2019 held in Kolkata, India.

The mediated role of complement C3 in PM2.5 exposure and type 2 diabetes mellitus: an elderly panel study in Beijing, China.

Diabetes mellitus (DM) is a common chronic disease worldwide. Ambient air pollution has long been proven to be associated with type 2 diabetes mellitus (T2DM) progression, but the underlying mechanism is not clear yet. In addition, previous studies mainly focused on the prevention of healthy people against the incidence of T2DM. We designed a panel study including two follow-ups and enrolled 39 patients with T2DM living in Beijing. Linear mixed model was fitted to assess the association between two pairs of variables (ambient air pollution exposure and C3 levels, ambient air pollution exposures and T2DM index). Mediation analysis of C3 between ambient air pollution exposure and indicators of T2DM progression was conducted. We found that PM2.5 exposures is are negatively associated with serum complement C3. Given that C3 might act as a protector of pancreas ß cell, PM2.5 exposures could accelerate disease in T2DM populations. No mediation effects were found. This study reveals that exposures to PM2.5 can cause progression of diseases among T2DM populations.

The Effect of Liposuction Cannula Diameter on Fat Retention-Based on a Rheological Simulation.

BACKGROUND: Autologous fat is considered as an ideal material for soft-tissue augmentation in plastic and reconstructive surgery. The primary drawback of autologous fat grafting is the high absorption rate, thus fat retention is considered as an essential indicator. There are several researches about the factors that can influence fat retention, including centrifugation and cannula size. However, rheological models of cannula during liposuction are limited. This research focuses on the effects of cannulas with diameters of 2 mm and 2.5 mm on fat retention, which is based on a rheological simulation of inlet pressure and maximum velocity. Experiments on mice were also conducted to confirm the result from the simulation. METHODS: A simulation was conducted with the physical parameters of the adipose tissue. Human lipoaspirate samples were obtained from patients by liposuction through cannulas of different diameters and were transferred into subcutaneous tissue of nude mice, a part of which were used in viscosity and density measurement. Graft retention was measured and fat quality was assessed through histologic analysis after 6 months. RESULTS: Viscosity and density of the fat tissue had significant effects on fat retention. The 2.5 mm diameter cannula had significantly lower inlet pressure and maximum velocity and thus led to higher graft retention, but oil cystic nodules appeared meanwhile. CONCLUSIONS: Cannulas with larger diameters have lower inlet pressure and maximum velocity during the liposuction process, which further influences the viability of adipocytes and adipose stem cells and thus has larger fat graft retention. This research built a mathematical model with less bias than in vivo experiments and provides a general way for analyzing the outcome of a liposuction precisely, which adds to the data for cannula optimization.