Learning the Protein Language Model of SARS-CoV-2 Spike Proteins

Ahstract-SARS-CoV-2 virus has long been evolving posing an increased risk in terms of infectivity and transmissibility which causes greater impact in communities worldwide. With the surge of collected SARS-CoV-2 sequences, studies found out that most of the emerging variants are linked to increased mutations in the spike (S) protein as observed in Alpha, Beta, Gamma, and Delta variants. Multiple approaches on genomic surveillance have been performed to monitor the mutational status and spread of the virus however most are heavily dependent on labels attributed to these sequences. Hence, this study features a system that has the capability to learn the protein language model of SARS-CoV-2 spike proteins, based on a bidirectional long-short term memory (BiLSTM) recurrent neural network, using sequence data alone. Upon obtaining the sequence embedding from the model, observed clusters are generated using the Leiden clustering algorithm and is visualized to monitor similarities between variants in terms of grammatical probability and semantic change. Additionally, the system measures the validity of a user-generated next-generation sequence capturing potential sequence mutations indicative of viral escape, particularly mutations by substitutions. Further studies on methods uncovering semantic rules that govern spike proteins are recommended to learn more about other viral characteristics conclusive of the future of the COVID-19 pandemic. © 2023 IEEE.

A Floyd-Warshall-based Reoptimization of Q Matrix on the Single DVRPPD with On-demand Cancellations

Delivery services have reached an all-time high in terms of demands due to the COVID-19 pandemic. An optimal routing plan for these different courier services is a must. The vehicle routing problem (VRP) is an NP-hard problem in logistics research and it can be solved by exact algorithms, heuristics, and machine learning through reinforcement learning. This study introduces a reoptimization alternative as opposed to trivial Q-learning retraining on a newly introduced subproblem of dynamic VRP and VRPPD, the dynamic vehicle routing problem with pickup, delivery, and cancellation (DVRPPDC). The reoptimization technique is called Floyd-Warshall LookUp Reoptimization of Rewards Yearned (FLURRY). Combined with a one-time Q-learning computation beforehand, the Q matrix produced is updated at the cell level by a lookup table containing all the shortest paths among the pairs of parcel lockers. The lookup table is generated via Floyd-Warshall algorithm, a well-known shortest path algorithm. Upon testing on a new dataset for DVRPPDC, one-time Q-learning combined with FLURRY is 6.1x to 10.6x faster to compute than Q-learning retraining. Furthermore, an additional study done after the main series of experiments reveal that the methodology does not necessitate any Q-learning training at the first time step at all. FLURRY can be done on a Q matrix of zeroes and achieve the same path output and traversal time as Q-learning with FLURRY. The standalone FLURRY algorithm further speeds up the computation, compared to the naive Q-learning approach, from 6.1x - 10.6x to 26.5x - 117.5x.

A Machine Learning Approach in Evaluating Symptom Screening in Predicting COVID-19

COVID-19 is a disease caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) that, to date, has over 245 million confirmed cases and claimed almost 5 million lives. This disease attacks the respiratory system and comes with a number of symptoms. The US Center for Disease Control and Prevention presents a set of symptoms. However, these symptoms only begin to manifest after a number of days, which prevents early detection of this disease. This absence of symptoms during the early stages is what is considered by many to be the very factor that caused the virus into becoming a pandemic. Nonetheless, symptoms checking has been used in practice by commercial and business establishments as an initial screening for COVID-19. The bothersome process of symptom checking are still in place at the entrances of malls and airports. In this study, we determine whether or not symptom screening is an effective system to be employed to assess individuals for COVID-19. Specifically, it aims to determine whether or not one or a set of symptoms are effective predictors of the RT-PCR test results, the gold standard in Covid-19 testing, using machine learning. Using data from the Philippine Red Cross, classification models are developed using LightGBM, AdaBoost, Gaussian Naïve-Bayes, MultiLayer Perceptron, Quadratic Discriminant Analysis and Decision Tree. These models were evaluated using the following metrics: precision, sensitivity, specificity and the type II error rate. Furthermore, for explainability, symptoms are analyzed as to whether or not they are relatively influential on the predicting whether or not a patient has COVID-19. The high type II error rate, low sensitivity and low relative predictor scores of the most significant predictor symptoms clearly show that symptoms do not correlate with the RT-PCR testing results. Thus, we conclude that symptom screening is not a medically suitable process for determining whether an individual has COVID-19. In fact, it even exposes us to the risk of viral transmission as people congregate at the entrances and lobbies of establishments. © 2022 IEEE.

Identification and Analysis of COVID-19-related Misinformation Tweets via Kullback-Leibler Divergence for Informativeness and Phraseness and Biterm Topic Modeling

The interaction of Filipinos transitioned to a virtual setting making social media, like Twitter, their source of information since the pandemic started. The infodemic it caused has opened up avenues to understand the characteristics of misinformation tweets regarding COVID-19. In this paper, we present the classification and analysis of misinformation tweets related to COVID-19 towards identifying themes. We used pointwise KL divergence in scoring "informativeness"and "phraseness"to extract misinformation tweets and BTM for topic modeling. With a testbed of 7, 711 tweets, the classifier model identified 3, 533 misinformation tweets with an accuracy of 74.25%. The results of the topic modeling were analyzed and clustered to expose possible narratives in the data set. The three narratives showed that most Filipinos use Twitter to share jokes, spread information and awareness about the virus, express opinions about the government's response, and share tips to prevent the disease. A wider date coverage could be included in future works. © 2022 IEEE.

Predicting your next trip: A knowledge graph-based multi-task learning approach for travel destination recommendation

Recommending the next destination to a traveler is a task that has been at the forefront of the airline industry for a long time, and its relevance has never been more important than today to revive tourism after the Covid-19 crisis. Several factors influence a user's decision when faced with a variety of travel destination choices: geographic context, best time to go, personal experiences, places to visit, scheduled events, etc. The challenge of recommending the right travel destination lies in efficiently integrating and leveraging all of this information into the recommender system. Based on a real world application scenario, we propose a multi-task learning model based on a neural network architecture that leverages knowledge graph to recommend the next destination to a traveler. We experimentally evaluated our proposed approach by comparing it against the currently in-production system and state-of-the-art travel destination recommendation algorithms in an offline setting. The results confirm the significant contribution of using knowledge graphs as a means of representing the heterogeneous information used for the recommendation task, as well as the benefit of using a multi-task learning model in terms of recommendation performance and training time. Copyright © 2021 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).

Determining factors contributing to the psychological impact of the COVID-19 Pandemic using machine learning

The COVID-19 pandemic has created a massive impact in the economy, healthcare, education and other aspects of society in each and every respect. Also greatly affected is the mental health of individuals. This study aims to determine the possible contributing factors to stress, anxiety, depression, and adverse psychological impact on the general population of the Philippines using machine learning approaches. The data gathered from 2119 participants who answered an online survey was analyzed using feature selection methods and machine learning classifiers to determine contributing factors to the aforementioned mental health issues. The results show that longer hours at home, on social media, age, how people rate their own health, pre-existence of a neuropsychiatric condition;wanting information on availability and effectiveness of a medicine or vaccine. being concerned for their family., feeling discriminated;and symptoms of body pain, difficulty breathing. and cough were good predictors of individuals being adversely impacted psychologically by the pandemic and others having elevated levels of stress, anxiety, depression.