Using artificial intelligence to improve human performance: efficient retinal disease detection training with synthetic images.

BACKGROUND: Artificial intelligence (AI) in medical imaging diagnostics has huge potential, but human judgement is still indispensable. We propose an AI-aided teaching method that leverages generative AI to train students on many images while preserving patient privacy. METHODS: A web-based course was designed using 600 synthetic ultra-widefield (UWF) retinal images to teach students to detect disease in these images. The images were generated by stable diffusion, a large generative foundation model, which we fine-tuned with 6285 real UWF images from six categories: five retinal diseases (age-related macular degeneration, glaucoma, diabetic retinopathy, retinal detachment and retinal vein occlusion) and normal. 161 trainee orthoptists took the course. They were evaluated with two tests: one consisting of UWF images and another of standard field (SF) images, which the students had not encountered in the course. Both tests contained 120 real patient images, 20 per category. The students took both tests once before and after training, with a cool-off period in between. RESULTS: On average, students completed the course in 53 min, significantly improving their diagnostic accuracy. For UWF images, student accuracy increased from 43.6% to 74.1% (p<0.0001 by paired t-test), nearly matching the previously published state-of-the-art AI model's accuracy of 73.3%. For SF images, student accuracy rose from 42.7% to 68.7% (p<0.0001), surpassing the state-of-the-art AI model's 40%. CONCLUSION: Synthetic images can be used effectively in medical education. We also found that humans are more robust to novel situations than AI models, thus showcasing human judgement's essential role in medical diagnosis.

Foveal Thickness Fluctuations in Anti-VEGF Treatment for Central Retinal Vein Occlusion.

Purpose: The aim of this study was to examine the effects of foveal thickness (FT) fluctuation (FTF) on 2-year visual and morphological outcomes of eyes with central retinal vein occlusion (CRVO) undergoing anti-VEGF treatment for recurrent macular edema (ME) based on a pro re nata regimen. Design: Retrospective, observational case series. Participants: We analyzed 141 treatment-naive patients (141 eyes) with CRVO-ME at a multicenter retinal practice. Methods: We assessed FT using OCT at each study visit. Patients were divided into groups 0, 1, 2, and 3 according to increasing FTF. Main Outcome Measures: We evaluated the logarithm of the minimal angle of resolution (logMAR) best-corrected visual acuity (BCVA), the length of the foveal ellipsoid zone (EZ) band defect measured using OCT, and the association of FTF with VA and EZ band defect length. Results: The mean baseline logMAR BCVA and FT were 0.65 ± 0.52 (Snellen equivalent range: 20/20-20/2000) and 661.1 ± 257.4 µm, respectively. The mean number of anti-VEGF injections administered was 5.6 ± 3.6. At the final examination, the mean logMAR BCVA and FT values were significantly improved relative to the baseline values (both P < 0.01). During the observation, BCVA longitudinally improved in Groups 0 and 1, remained unchanged in Group 2, and worsened in Group 3. Likewise, the length of the foveal EZ band defect did not increase in Group 0; however, it gradually increased in Groups 1, 2, and 3. Foveal thickness fluctuation was significantly and positively associated with the logMAR BCVA and length of the foveal EZ band defect at the final examination (P < 0.01). The final logMAR BCVA of patients developing neovascular complications was 1.27 ± 0.72 (Snellen equivalent range: 20/50-counting fingers), which was significantly poorer than that of patients without complications (P < 0.001). There was no significant difference in the neovascular complication rate among the FTF groups (P = 0.106, Fisher exact test). Conclusions: In eyes receiving anti-VEGF treatment for CRVO-ME, FTF can longitudinally impair the visual acuity and foveal photoreceptor status during the observation period, thus influencing the final outcomes. However, neovascular complications, which would also lead to a poor visual prognosis, may not be associated with FTF. Financial Disclosures: The authors have no proprietary or commercial interest in any materials discussed in this article.

Revolutionizing Patient Monitoring in Age-Related Macular Degeneration: A Comparative Study on the Necessity and Efficiency of the AMD VIEWER.

(1) Background: Age-related Macular Degeneration (AMD) is a critical condition leading to blindness, necessitating lifelong clinic visits for management, albeit with existing challenges in monitoring its long-term progression. This study introduced and assessed an innovative tool, the AMD long-term Information Viewer (AMD VIEWER), designed to offer a comprehensive display of crucial medical data-including visual acuity, central retinal thickness, macular volume, vitreous injection treatment history, and Optical Coherent Tomography (OCT) images-across an individual eye's entire treatment course. (2) Methods: By analyzing visit frequencies of patients with a history of invasive AMD treatment, a comparative examination between a Dropout group and an Active group underscored the clinical importance of regular visits, particularly highlighting better treatment outcomes and maintained visual acuity in the Active group. (3) Results: The efficiency of AMD VIEWER was proven by comparing it to manual data input by optometrists, showing significantly faster data display with no errors, unlike the time-consuming and error-prone manual entries. Furthermore, an elicited Net Promoter Score (NPS) of 70 from 10 ophthalmologists strongly endorsed AMD VIEWER's practical utility. (4) Conclusions: This study underscores the importance of regular clinic visits for AMD patients. It suggests the AMD VIEWER as an effective tool for improving treatment data management and display.

A novel in silico scaffold-hopping method for drug repositioning in rare and intractable diseases.

In the field of rare and intractable diseases, new drug development is difficult and drug repositioning (DR) is a key method to improve this situation. In this study, we present a new method for finding DR candidates utilizing virtual screening, which integrates amino acid interaction mapping into scaffold-hopping (AI-AAM). At first, we used a spleen associated tyrosine kinase inhibitor as a reference to evaluate the technique, and succeeded in scaffold-hopping maintaining the pharmacological activity. Then we applied this method to five drugs and obtained 144 compounds with diverse structures. Among these, 31 compounds were known to target the same proteins as their reference compounds and 113 compounds were known to target different proteins. We found that AI-AAM dominantly selected functionally similar compounds; thus, these selected compounds may represent improved alternatives to their reference compounds. Moreover, the latter compounds were presumed to bind to the targets of their references as well. This new "compound-target" information provided DR candidates that could be utilized for future drug development.

Training data size and predication errors in the use of machine-learning assisted intraocular lens power calculation.

This retrospective study examined the effect of the size of training data on the accuracy of machine learning-assisted SRK/T power calculation. Clinical records of 4800 eyes of 4800 Japanese patients with intraocular lenses (IOLs) were reviewed. A support vector regressor (SVR) was used for refining the SRK/T formula, and dataset sizes for training and evaluation were reduced from full to 1/64. The prediction errors from the postoperative refractions were calculated, and the proportion within ± 0.25 D, ± 0.50 D, and ± 1.00 D of errors were compared with those using full data. The influence of the difference in A-constant was also evaluated. Prediction errors within ± 0.50 D in the use of full data were obtained with the dataset of ≥ 150 eyes (P = 0.016), whereas the datasets of ≥ 300 eyes were required for the error within ± 0.25 D (P < 0.030). The prediction errors did not alter with the A-constant values among IOLs with open-loop haptics, except for IOLs with plated haptics. In conclusion, the accuracy of SVR-assisted SRK/T could be achieved with the training dataset of ≥ 150 eyes for the Japanese population, and the calculation was versatile for any open-looped IOLs.

Estimation of Visual Function Using Deep Learning From Ultra-Widefield Fundus Images of Eyes With Retinitis Pigmentosa.

Importance: There is no widespread effective treatment to halt the progression of retinitis pigmentosa. Consequently, adequate assessment and estimation of residual visual function are important clinically. Objective: To examine whether deep learning can accurately estimate the visual function of patients with retinitis pigmentosa by using ultra-widefield fundus images obtained on concurrent visits. Design, Setting, and Participants: Data for this multicenter, retrospective, cross-sectional study were collected between January 1, 2012, and December 31, 2018. This study included 695 consecutive patients with retinitis pigmentosa who were examined at 5 institutions. Each of the 3 types of input images-ultra-widefield pseudocolor images, ultra-widefield fundus autofluorescence images, and both ultra-widefield pseudocolor and fundus autofluorescence images-was paired with 1 of the 31 types of ensemble models constructed from 5 deep learning models (Visual Geometry Group-16, Residual Network-50, InceptionV3, DenseNet121, and EfficientNetB0). We used 848, 212, and 214 images for the training, validation, and testing data, respectively. All data from 1 institution were used for the independent testing data. Data analysis was performed from June 7, 2021, to December 5, 2022. Main Outcomes and Measures: The mean deviation on the Humphrey field analyzer, central retinal sensitivity, and best-corrected visual acuity were estimated. The image type-ensemble model combination that yielded the smallest mean absolute error was defined as the model with the best estimation accuracy. After removal of the bias of including both eyes with the generalized linear mixed model, correlations between the actual values of the testing data and the estimated values by the best accuracy model were examined by calculating standardized regression coefficients and P values. Results: The study included 1274 eyes of 695 patients. A total of 385 patients were female (55.4%), and the mean (SD) age was 53.9 (17.2) years. Among the 3 types of images, the model using ultra-widefield fundus autofluorescence images alone provided the best estimation accuracy for mean deviation, central sensitivity, and visual acuity. Standardized regression coefficients were 0.684 (95% CI, 0.567-0.802) for the mean deviation estimation, 0.697 (95% CI, 0.590-0.804) for the central sensitivity estimation, and 0.309 (95% CI, 0.187-0.430) for the visual acuity estimation (all P < .001). Conclusions and Relevance: Results of this study suggest that the visual function estimation in patients with retinitis pigmentosa from ultra-widefield fundus autofluorescence images using deep learning might help assess disease progression objectively. Findings also suggest that deep learning models might monitor the progression of retinitis pigmentosa efficiently during follow-up.

Transferability of an Artificial Intelligence Algorithm Predicting Rebubblings After Descemet Membrane Endothelial Keratoplasty.

PURPOSE: To develop an artificial intelligence (AI) algorithm enabling corneal surgeons to predict the probability of rebubbling after Descemet membrane endothelial keratoplasty (DMEK) from images obtained using optical coherence tomography (OCT). METHODS: Anterior segment OCT data of patients undergoing DMEK by 2 different DMEK surgeons (C.C. and B.B.; University of Cologne, Cologne, Germany) were extracted from the prospective Cologne DMEK database. An AI algorithm was trained by using a data set of C.C. to detect graft detachments and predict the probability of a rebubbling. The architecture of the AI model used in this study was called EfficientNet. This algorithm was applied to OCT scans of patients, which were operated by B.B. The transferability of this algorithm was analyzed to predict a rebubbling after DMEK. RESULTS: The algorithm reached an area under the curve of 0.875 (95% confidence interval: 0.880-0.929). The cutoff value based on the Youden index was 0.214, and the sensitivity and specificity for this value were 78.9% (67.6%-87.7%) and 78.6% (69.5%-86.1%). CONCLUSIONS: The development of AI algorithms allows good transferability to other surgeons reaching a high accuracy in predicting rebubbling after DMEK based on OCT image data.

Automatic Diagnosis of Diabetic Retinopathy Stage Focusing Exclusively on Retinal Hemorrhage.

Background and Objectives: The present study evaluated the detection of diabetic retinopathy (DR) using an automated fundus camera focusing exclusively on retinal hemorrhage (RH) using a deep convolutional neural network, which is a machine-learning technology. Materials and Methods: This investigation was conducted via a prospective and observational study. The study included 89 fundus ophthalmoscopy images. Seventy images passed an image quality review and were graded as showing no apparent DR (n = 51), mild nonproliferative DR (NPDR; n = 16), moderate NPDR (n = 1), severe NPDR (n = 1), and proliferative DR (n = 1) by three retinal experts according to the International Clinical Diabetic Retinopathy Severity scale. The RH numbers and areas were automatically detected and the results of two tests-the detection of mild-or-worse NPDR and the detection of moderate-or-worse NPDR-were examined. Results: The detection of mild-or-worse DR showed a sensitivity of 0.812 (95% confidence interval: 0.680-0.945), specificity of 0.888, and area under the curve (AUC) of 0.884, whereas the detection of moderate-or-worse DR showed a sensitivity of 1.0, specificity of 1.0, and AUC of 1.0. Conclusions: Automated diagnosis using artificial intelligence focusing exclusively on RH could be used to diagnose DR requiring ophthalmologist intervention.

Automated detection of enlarged extraocular muscle in Graves' ophthalmopathy with computed tomography and deep neural network.

This study aimed to develop a diagnostic software system to evaluate the enlarged extraocular muscles (EEM) in patients with Graves' ophthalmopathy (GO) by a deep neural network.This prospective observational study involved 371 participants (199 EEM patients with GO and 172 controls with normal extraocular muscles) whose extraocular muscles were examined with orbital coronal computed tomography. When at least one rectus muscle (right or left superior, inferior, medial, or lateral) in the patients was 4.0 mm or larger, it was classified as an EEM patient with GO. We used 222 images of the data from patients as the training data, 74 images as the validation test data, and 75 images as the test data to "train" the deep neural network to judge the thickness of the extraocular muscles on computed tomography. We then validated the performance of the network. In the test data, the area under the curve was 0.946 (95% confidence interval (CI) 0.894-0.998), and receiver operating characteristic analysis demonstrated 92.5% (95% CI 0.796-0.984) sensitivity and 88.6% (95% CI 0.733-0.968) specificity. The results suggest that the deep learning system with the deep neural network can detect EEM in patients with GO.

Deep Learning-Based System for Preoperative Safety Management in Cataract Surgery.

An artificial intelligence-based system was implemented for preoperative safety management in cataract surgery, including facial recognition, laterality (right and left eye) confirmation, and intraocular lens (IOL) parameter verification. A deep-learning model was constructed with a face identification development kit for facial recognition, the You Only Look Once Version 3 (YOLOv3) algorithm for laterality confirmation, and the Visual Geometry Group-16 (VGG-16) for IOL parameter verification. In 171 patients who were undergoing phacoemulsification and IOL implantation, a mobile device (iPad mini, Apple Inc.) camera was used to capture patients' faces, location of surgical drape aperture, and IOL parameter descriptions on the packages, which were then checked with the information stored in the referral database. The authentication rates on the first attempt and after repeated attempts were 92.0% and 96.3% for facial recognition, 82.5% and 98.2% for laterality confirmation, and 67.4% and 88.9% for IOL parameter verification, respectively. After authentication, both the false rejection rate and the false acceptance rate were 0% for all three parameters. An artificial intelligence-based system for preoperative safety management was implemented in real cataract surgery with a passable authentication rate and very high accuracy.

Foveal Thickness Fluctuation in Anti-VEGF Treatment for Branch Retinal Vein Occlusion: A Long-term Study.

PURPOSE: Branch retinal vein occlusion (BRVO) causes macular edema (ME), which can be controlled with anti-VEGF treatments. However, these treatments are not curative, necessitating additional anti-VEGF treatments at recurrence. Long-term results, optimal anti-VEGF treatment regimens, and the comprehensive effects of ME recurrence are largely unknown. Thus, we aimed to examine the effects of foveal thickness (FT) fluctuation (FTF) on the visual and morphologic outcomes of anti-VEGF treatments for BRVO-ME administered via a pro re nata regimen. DESIGN: A retrospective, observational case series. SUBJECTS: This study analyzed 309 treatment-naïve patients (309 eyes) with BRVO-ME between 2012 and 2021 at a multicenter retinal practice. METHODS: The FT was assessed using OCT at each study visit. MAIN OUTCOME MEASURES: We evaluated the logarithm of the minimal angle of resolution (logMAR) best corrected visual acuity (BCVA) and the defect length of the foveal ellipsoid zone (EZ) band using OCT. RESULTS: At baseline, the mean logMAR BCVA was 0.30 ± 0.30 and the mean FT was 503 ± 162 µm. The number of anti-VEGF injections for BRVO-ME was 5.8 ± 4.6 during the mean follow-up period (50.6 ± 22.2 months). At the final examination, the mean logMAR BCVA and FT values were significantly improved compared with those at the baseline. Multiple regression analyses showed that age, baseline logMAR BCVA, and FTF were significantly associated with the final logMAR BCVA (ß = 0.20, 0.35, and 0.30, respectively). Foveal thickness fluctuation (divided into groups 0-3 in ascending order of FTF) was significantly associated with logMAR BCVA and the defect length of the foveal EZ band at the final examination. The defect lengths of the foveal EZ band were longitudinally shortened in groups 0 and 1 and were slightly prolonged in groups 2 and 3. The logMAR BCVA showed improvements in groups 0 and 1 and worsened slightly in groups 2 and 3. CONCLUSIONS: Foveal thickness fluctuation was significantly associated with visual acuity and foveal photoreceptor status. Thus, the morphologic and functional prognoses of eyes with BRVO may improve with the identification of the characteristics of eyes with greater FTF and consequently controlling the FTF more strictly.

Nationwide multicentre comparison of preoperative biometry and predictability of cataract surgery in Japan.

AIM: To compare the preoperative biometric data and the refractive accuracy of cataract surgery among major surgical sites in a nationwide multicentre study. METHODS: We prospectively obtained the preoperative biometric data of 2143 eyes of 2143 consecutive patients undergoing standard cataract surgery at major 12 facilities and compared the preoperative biometry as well as the postoperative refractive accuracy among them. RESULTS: We found significant differences in most preoperative variables, such as axial length (one-way analysis of variance, p=0.003), anterior chamber depth (p<0.001), lens thickness (p<0.001) and central corneal thickness (p<0.001), except for mean keratometry (p=0.587) and corneal astigmatism (p=0.304), among the 12 surgical sites. The prediction error using the Sanders-Retzlaff-Kraff/Theoretical (SRK/T formula was significantly more hyperopic than that using the Barrett Universal II formula (paired t-test, p<0.001). The absolute error using the SRK/T formula was significantly larger than that using the Barrett Universal II formula (p=0.016). The prediction error using the SRK/T formula was significantly more hyperopic than that using the Barrett Universal II formula at 10 of 12 institutions, but significantly more myopic at one institution. The absolute error using the SRK/T formula was significantly larger than that using the Barrett Universal II formula at 4 of 12 institutions but significantly smaller at two institutions. CONCLUSIONS: Regional divergences of the preoperative biometry were not necessarily negligible, and the optimised intraocular lens power calculation formula was individually different among the 12 facilities. Our findings highlight the importance of individual optimisation of these formulas at each facility, especially in consideration of these biometric variations.Trial registration numberClinical Trial Registry; 000039976.

Developing an iOS application that uses machine learning for the automated diagnosis of blepharoptosis.

PURPOSE: To assess the performance of artificial intelligence in the automated classification of images taken with a tablet device of patients with blepharoptosis and subjects with normal eyelid. METHODS: This is a prospective and observational study. A total of 1276 eyelid images (624 images from 347 blepharoptosis cases and 652 images from 367 normal controls) from 606 participants were analyzed. In order to obtain a sufficient number of images for analysis, 1 to 4 eyelid images were obtained from each participant. We developed a model by fully retraining the pre-trained MobileNetV2 convolutional neural network. Subsequently, we verified whether the automatic diagnosis of blepharoptosis was possible using the images. In addition, we visualized how the model captured the features of the test data with Score-CAM. k-fold cross-validation (k = 5) was adopted for splitting the training and validation. Sensitivity, specificity, and the area under the curve (AUC) of the receiver operating characteristic curve for detecting blepharoptosis were examined. RESULTS: We found the model had a sensitivity of 83.0% (95% confidence interval [CI], 79.8-85.9) and a specificity of 82.5% (95% CI, 79.4-85.4). The accuracy of the validation data was 82.8%, and the AUC was 0.900 (95% CI, 0.882-0.917). CONCLUSION: Artificial intelligence was able to classify with high accuracy images of blepharoptosis and normal eyelids taken using a tablet device. Thus, the diagnosis of blepharoptosis with a tablet device is possible at a high level of accuracy. TRIAL REGISTRATION: Date of registration: 2021-06-25. TRIAL REGISTRATION NUMBER: UMIN000044660. Registration site: https://upload.umin.ac.jp/cgi-open-bin/ctr/ctr_view.cgi?recptno=R000051004.

A deep learning approach for successful big-bubble formation prediction in deep anterior lamellar keratoplasty.

The efficacy of deep learning in predicting successful big-bubble (SBB) formation during deep anterior lamellar keratoplasty (DALK) was evaluated. Medical records of patients undergoing DALK at the University of Cologne, Germany between March 2013 and July 2019 were retrospectively analyzed. Patients were divided into two groups: (1) SBB or (2) failed big-bubble (FBB). Preoperative images of anterior segment optical coherence tomography and corneal biometric values (corneal thickness, corneal curvature, and densitometry) were evaluated. A deep neural network model, Visual Geometry Group-16, was selected to test the validation data, evaluate the model, create a heat map image, and calculate the area under the curve (AUC). This pilot study included 46 patients overall (11 women, 35 men). SBBs were more common in keratoconus eyes (KC eyes) than in corneal opacifications of other etiologies (non KC eyes) (p = 0.006). The AUC was 0.746 (95% confidence interval [CI] 0.603-0.889). The determination success rate was 78.3% (18/23 eyes) (95% CI 56.3-92.5%) for SBB and 69.6% (16/23 eyes) (95% CI 47.1-86.8%) for FBB. This automated system demonstrates the potential of SBB prediction in DALK. Although KC eyes had a higher SBB rate, no other specific findings were found in the corneal biometric data.

Predicting Keratoconus Progression and Need for Corneal Crosslinking Using Deep Learning.

We aimed to predict keratoconus progression and the need for corneal crosslinking (CXL) using deep learning (DL). Two hundred and seventy-four corneal tomography images taken by Pentacam HR® (Oculus, Wetzlar, Germany) of 158 keratoconus patients were examined. All patients were examined two times or more, and divided into two groups; the progression group and the non-progression group. An axial map of the frontal corneal plane, a pachymetry map, and a combination of these two maps at the initial examination were assessed according to the patients' age. Training with a convolutional neural network on these learning data objects was conducted. Ninety eyes showed progression and 184 eyes showed no progression. The axial map, the pachymetry map, and their combination combined with patients' age showed mean AUC values of 0.783, 0.784, and 0.814 (95% confidence interval (0.721-0.845) (0.722-0.846), and (0.755-0.872), respectively), with sensitivities of 87.8%, 77.8%, and 77.8% ((79.2-93.7), (67.8-85.9), and (67.8-85.9)) and specificities of 59.8%, 65.8%, and 69.6% ((52.3-66.9), (58.4-72.6), and (62.4-76.1)), respectively. Using the proposed DL neural network model, keratoconus progression can be predicted on corneal tomography maps combined with patients' age.

KEGG: integrating viruses and cellular organisms.

KEGG (https://www.kegg.jp/) is a manually curated resource integrating eighteen databases categorized into systems, genomic, chemical and health information. It also provides KEGG mapping tools, which enable understanding of cellular and organism-level functions from genome sequences and other molecular datasets. KEGG mapping is a predictive method of reconstructing molecular network systems from molecular building blocks based on the concept of functional orthologs. Since the introduction of the KEGG NETWORK database, various diseases have been associated with network variants, which are perturbed molecular networks caused by human gene variants, viruses, other pathogens and environmental factors. The network variation maps are created as aligned sets of related networks showing, for example, how different viruses inhibit or activate specific cellular signaling pathways. The KEGG pathway maps are now integrated with network variation maps in the NETWORK database, as well as with conserved functional units of KEGG modules and reaction modules in the MODULE database. The KO database for functional orthologs continues to be improved and virus KOs are being expanded for better understanding of virus-cell interactions and for enabling prediction of viral perturbations.

New approach for understanding genome variations in KEGG.

KEGG (Kyoto Encyclopedia of Genes and Genomes; https://www.kegg.jp/ or https://www.genome.jp/kegg/) is a reference knowledge base for biological interpretation of genome sequences and other high-throughput data. It is an integrated database consisting of three generic categories of systems information, genomic information and chemical information, and an additional human-specific category of health information. KEGG pathway maps, BRITE hierarchies and KEGG modules have been developed as generic molecular networks with KEGG Orthology nodes of functional orthologs so that KEGG pathway mapping and other procedures can be applied to any cellular organism. Unfortunately, however, this generic approach was inadequate for knowledge representation in the health information category, where variations of human genomes, especially disease-related variations, had to be considered. Thus, we have introduced a new approach where human gene variants are explicitly incorporated into what we call 'network variants' in the recently released KEGG NETWORK database. This allows accumulation of knowledge about disease-related perturbed molecular networks caused not only by gene variants, but also by viruses and other pathogens, environmental factors and drugs. We expect that KEGG NETWORK will become another reference knowledge base for the basic understanding of disease mechanisms and practical use in clinical sequencing and drug development.

KEGG: new perspectives on genomes, pathways, diseases and drugs.

KEGG (http://www.kegg.jp/ or http://www.genome.jp/kegg/) is an encyclopedia of genes and genomes. Assigning functional meanings to genes and genomes both at the molecular and higher levels is the primary objective of the KEGG database project. Molecular-level functions are stored in the KO (KEGG Orthology) database, where each KO is defined as a functional ortholog of genes and proteins. Higher-level functions are represented by networks of molecular interactions, reactions and relations in the forms of KEGG pathway maps, BRITE hierarchies and KEGG modules. In the past the KO database was developed for the purpose of defining nodes of molecular networks, but now the content has been expanded and the quality improved irrespective of whether or not the KOs appear in the three molecular network databases. The newly introduced addendum category of the GENES database is a collection of individual proteins whose functions are experimentally characterized and from which an increasing number of KOs are defined. Furthermore, the DISEASE and DRUG databases have been improved by systematic analysis of drug labels for better integration of diseases and drugs with the KEGG molecular networks. KEGG is moving towards becoming a comprehensive knowledge base for both functional interpretation and practical application of genomic information.

KEGG as a reference resource for gene and protein annotation.

KEGG (http://www.kegg.jp/ or http://www.genome.jp/kegg/) is an integrated database resource for biological interpretation of genome sequences and other high-throughput data. Molecular functions of genes and proteins are associated with ortholog groups and stored in the KEGG Orthology (KO) database. The KEGG pathway maps, BRITE hierarchies and KEGG modules are developed as networks of KO nodes, representing high-level functions of the cell and the organism. Currently, more than 4000 complete genomes are annotated with KOs in the KEGG GENES database, which can be used as a reference data set for KO assignment and subsequent reconstruction of KEGG pathways and other molecular networks. As an annotation resource, the following improvements have been made. First, each KO record is re-examined and associated with protein sequence data used in experiments of functional characterization. Second, the GENES database now includes viruses, plasmids, and the addendum category for functionally characterized proteins that are not represented in complete genomes. Third, new automatic annotation servers, BlastKOALA and GhostKOALA, are made available utilizing the non-redundant pangenome data set generated from the GENES database. As a resource for translational bioinformatics, various data sets are created for antimicrobial resistance and drug interaction networks.

Data, information, knowledge and principle: back to metabolism in KEGG.

In the hierarchy of data, information and knowledge, computational methods play a major role in the initial processing of data to extract information, but they alone become less effective to compile knowledge from information. The Kyoto Encyclopedia of Genes and Genomes (KEGG) resource (http://www.kegg.jp/ or http://www.genome.jp/kegg/) has been developed as a reference knowledge base to assist this latter process. In particular, the KEGG pathway maps are widely used for biological interpretation of genome sequences and other high-throughput data. The link from genomes to pathways is made through the KEGG Orthology system, a collection of manually defined ortholog groups identified by K numbers. To better automate this interpretation process the KEGG modules defined by Boolean expressions of K numbers have been expanded and improved. Once genes in a genome are annotated with K numbers, the KEGG modules can be computationally evaluated revealing metabolic capacities and other phenotypic features. The reaction modules, which represent chemical units of reactions, have been used to analyze design principles of metabolic networks and also to improve the definition of K numbers and associated annotations. For translational bioinformatics, the KEGG MEDICUS resource has been developed by integrating drug labels (package inserts) used in society.