• Deep neural networks-based predictors have enhanced the PSSP accuracy. • There is still an accuracy gap in PSSP, the PSSP theoretical limit expands to 94%. • The use of pre-trained LMs as input has successfully improved PSSP performance. • PSSP without using evolutionary information is a potential future work.

Vehicle make and model classification is crucial to the operation of an intelligent transportation system (ITS). Fine-grained vehicle information such as make and model can help officers uncover cases of traffic violations when license plate information cannot be obtained. Various techniques have been developed to perform vehicle make and model classification. However, it is very hard to identify the make and model of vehicles with highly similar visual appearances. The classifier contains a lot of potential for mistakes because the vehicles look very similar but have different models and manufacturers. To solve this problem, a fine-grained classifier based on convolutional neural networks with a multi-task learning approach is proposed in this paper. The proposed method takes a vehicle image as input and extracts features using the VGG-16 architecture. The extracted features will then be sent to two different branches, with one branch being used to classify the vehicle model and the other to classify the vehicle make. The performance of the proposed method was evaluated using the InaV-Dash dataset, which contains an Indonesian vehicle model with a highly similar visual appearance. The experimental results show that the proposed method achieves 98.73% accuracy for vehicle make and 97.69% accuracy for vehicle model. Our study also demonstrates that the proposed method is able to improve the performance of the baseline method on highly similar vehicle classification problems.

Comet assay is a simple and precise method to analyze DNA damage. Nowadays, many research studies have demonstrated the effectiveness of buccal mucosa cells usage in comet assays. However, several software tools do not perform well for detecting and classifying comets from a comet assay image of buccal mucosa cells because the cell has a lot more noise. Therefore, a specific software tool is required for fully automated comet detection and classification from buccal mucosa cell swabs. This research proposes a deep learning-based fully automated framework using Faster R-CNN to detect and classify comets in a comet assay image taken from buccal mucosa swab. To train the Faster R-CNN model, buccal mucosa samples were collected from 24 patients in Indonesia. We acquired 275 comet assay images containing 519 comets. Furthermore, two strategies were used to overcome the lack of dataset problems during the model training, namely transfer learning and data augmentation. We implemented the proposed Faster R-CNN model as a web-based tool, GamaComet, that can be accessed freely for academic purposes. To test the GamaComet, buccal mucosa samples were collected from seven patients in Indonesia. We acquired 43 comet assay images containing 73 comets. GamaComet can give an accuracy of 81.34% for the detection task and an accuracy of 66.67% for the classification task. Furthermore, we also compared the performance of GamaComet with an existing free software tool for comet detection, OpenComet. The experiment results showed that GamaComet performed significantly better than OpenComet that could only give an accuracy of 11.5% for the comet detection task. Downstream analysis can be well conducted based on the detection and classification results from GamaComet. The analysis showed that patients owning comet assay images containing comets with class 3 and class 4 had a smoking habit, meaning they had more cells with a high level of DNA damage. Although GamaComet had a good performance, the performance for the classification task could still be improved. Therefore, it will be one of the future works for the research development of GamaComet.

Purpose: More and more data are stored in text form due to technological developments, making text data processing more difficult. It also causes problems in the text preprocessing algorithm, one of which is when two texts are identical, but are considered distinct by the algorithm. Therefore, it is necessary to normalize the text to get the standard form of words in a particular language. Spelling correction is often used to normalize text, but for Bahasa Indonesia, there has not been much research on the spell correction algorithm. Thus, there needs to be a comparison of the most appropriate spelling correction algorithms for the normalization process to be effective.Methods: In this study, we compared three algorithms, namely Levenshtein Distance, Jaro-Winkler Distance, and Smith-Waterman. These algorithms were evaluated using questionnaire data and tweet data, which both are in Bahasa Indonesia.Result: The fastest normalization time is obtained by the Jaro-Winkler, taking an average of 31.01 seconds for questionnaire data and 59.27 seconds for tweet data. The best accuracy is obtained by the Levenshtein Distance with a value of 44.90% for the questionnaire data and 60.04% for the tweet data. Novelty: The novelty of this research is to compare the similarity measure algorithm in Bahasa Indonesia. Therefore, the most suitable similarity measure algorithm for Bahasa Indonesia will be obtained.

Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) is a newly emerging virus well known as the major cause of the worldwide pandemic due to Coronavirus Disease 2019 (COVID-19). Major breakthroughs in the Next Generation Sequencing (NGS) field were elucidated following the first release of a full-length SARS-CoV-2 genome on the 10 January 2020, with the hope of turning the table against the worsening pandemic situation. Previous studies in respiratory virus characterization require mapping of raw sequences to the human genome in the downstream bioinformatics pipeline as part of metagenomic principles. Illumina, as the major player in the NGS arena, took action by releasing guidelines for improved enrichment kits called the Respiratory Virus Oligo Panel (RVOP) based on a hybridization capture method capable of capturing targeted respiratory viruses, including SARS-CoV-2; therefore, allowing a direct map of raw sequences data to SARS-CoV-2 genome in downstream bioinformatics pipeline. Consequently, two bioinformatics pipelines emerged with no previous studies benchmarking the pipelines. This study focuses on gaining insight and understanding of target enrichment workflow by Illumina through the utilization of different bioinformatics pipelines named as ‘Fast Pipeline’ and ‘Normal Pipeline’ to SARS-CoV-2 strains isolated from Yogyakarta and Central Java, Indonesia. Overall, both pipelines work well in the characterization of SARS-CoV-2 samples, including in the identification of major studied nucleotide substitutions and amino acid mutations. A higher number of reads mapped to the SARS-CoV-2 genome in Fast Pipeline and merely were discovered as a contributing factor in a higher number of coverage depth and identified variations (SNPs, insertion, and deletion). Fast Pipeline ultimately works well in a situation where time is a critical factor. On the other hand, Normal Pipeline would require a longer time as it mapped reads to the human genome. Certain limitations were identified in terms of pipeline algorithm, whereas it is highly recommended in future studies to design a pipeline in an integrated framework, for instance, by using NextFlow, a workflow framework to combine all scripts into one fully integrated pipeline.