To increase the success in Covid 19 treatment, many drug suggestions are presented, and some clinical studies are shared in the literature. There have been some attempts to use some of these drugs in combination. However, using more than one drug together may cause serious side effects on patients. Therefore, detecting drug-drug interactions of the drugs used will be of great importance in the treatment of Covid 19. In this study, the interactions of 8 drugs used for Covid 19 treatment with 645 different drugs and possible side effects estimates have been produced using Graph Convolutional Networks. Organ systems and diseases in which these 8 drugs cause the most negative effects have been identified. In addition, as it is known that some of these 8 drugs are used together in Covid-19 treatment, the side effects caused by using these drugs together are shared. With the experimental results obtained, it is aimed to facilitate the selection of the drugs and increase the success of Covid 19 treatment according to the targeted patient.
Method Article
Identifying Side Effects of Some Drugs Used in Covid-19 Treatment
https://doi.org/10.21203/rs.3.rs-40903/v1
This work is licensed under a CC BY 4.0 License
Version 1
posted 14 Jul, 2020
You are reading this latest preprint version
Covid 19
Graph Convolutional Networks
Drug Drug Interaction
Side Effects
Adverse Drug Reaction
As of December 2019, a coronavirus species that can spread from person to person was identified in Wuhan, China (F. Wu et al., 2020). The disease later on called Covid-19 posed a risk to be declared as a pandemic by the World Health Organization (WHO) in a short time (WHO, 2020). As of the date of this study in early June 2020, more than 6.4 million people were infected with this virus, and 373,334 persons died. It has been realized that scientists and researchers have published thousands of clinical trial results and articles in this process to provide treatment methods for the disease (Sanders et al., 2020). An important part of these studies examines the use of existing drugs for Covid-19 treatment, and suggests possible treatment methods, e.g., (Colson et al., 2020; Stebbing et al., 2020; Zhou et al., 2020).
One of the issues that should be examined before recommending a drug to a patient and after treatment is the side effects of the drug. Research shows that multiple drugs usage (polypharmacy) significantly increases drug side effects (Colley & Lucas, 1993; Guthrie et al., 2015). For older patients, the probability of polypharmacy generally increases. However, studies clearly show that as the number of drugs used increases, the negative effects seen in patients may also increase (Hajjar et al., 2007; Lavan & Gallagher, 2016; Tatum et al., 2019). Therefore, it is vital to predict drug-drug interactions (DDI) and adverse drug reactions (ADR) for the drugs to be used in the treatment of a disease (Colley & Lucas, 1993; Pirmohamed et al., 2004). Knowing the side effects and DDI of the drugs recommended in Covid-19 treatment will play an important role in the success of the process.
According to statistical studies, the vast majority of Covid-19 patients are seen at age 50 and over (Sobotka et al., 2020). According to the results of polypharmacy studies, regular and multiple drugs usage is over 60% in this age group (Guthrie et al., 2015). When these two examinations are evaluated together, it is understood that the rates of multiple drugs usage for Covid-19 patients are quite high. For this reason, the importance of DDI studies increases in Covid-19 treatment. Recognizing the drugs used regularly by the patient is a factor that will directly affect the selection of the drugs to be used in the treatment process. Many studies have showed that treatment success has increased significantly, thanks to patient-focused drug combinations (Makiani et al., 2017; W. Sun et al., 2016). The role of DDI studies to improve success in Covid-19 therapy is better understood with previous studies.
With advancing technology, alternative methods have been developed for clinical studies in DDI detection. Today, DDI research efforts are mainly conducted using computer-based methods (Zhao et al., 2010). In this paper, possible interactions of drugs used in Covid-19 treatment will be examined by employing graph convolutional networks. The aim of the study is to create a projection on the interactions of drugs used in Covid-19 treatment with other drugs. This way, it is aimed to contribute to increasing the success of the treatment by reducing the negative effects of the drugs. To achieve this, answers to the following questions were sought for each of the 8 drugs examined within the scope of the study.
- Which other drugs used in combination with a given drug will have the most dangerous side effects?
- What are the most common side effects as a result of using another drug with a given drug?
- Which disease or organ system are mostly affected by the side effects caused by the given drug?
This section concisely reviews previous research efforts described in the literature. Firstly, DDI studies for Covid-19 treatment are investigated. Then, DDI studies, and methods with different approaches were examined.
A. DDI Studies for the Treatment of COVID 19
Drug interactions for different drugs and disease groups were examined during the treatment of the disease. For example, it is recommended that the risks should be evaluated well before using Ritonavir / Lopinavir drugs in patients with kidney transplantation (Bartiromo et al., 2020). In addition, the need for guidelines to be prepared on this subject was emphasized. In a different study (Roden et al., 2020), possible effects of drugs were investigated according to the heart rhythm graphics of patients. This research shares the side effects of the combination of Hydroxychloroquine and Azithromycin when used together in Covid- 19 treatment. In a more comprehensive research, interactions of 4 different drug groups were examined (Elens et al., 2020). In addition to these studies, there are websites which are available for online access by universities and pharmaceutical research laboratories (COVID-19 Drug Information, 2020; Medscape Drug Reference Database, 2020; Liverpool COVID-19 Interactions, 2020).
B. Computer Based DDI Studies
Extracting complex relationships from big data has become possible with today's technology and data analysis methods. Since the concept of DDI focuses on the relationships between drugs, many studies have been conducted using data mining methods (Vilar et al., 2018). In such studies, a corpus with pharmacological data on drugs is usually used. The next step is to apply data mining methods to extract relationships from this corpus (Hammann & Drewe, 2014; Iyer et al., 2014; Lorberbaum et al., 2016).
Another method of relationship extraction using information technology employs neural networks (Niepert et al., 2016). These structures, consisting of nodes and edges, form the basis for studies that can reveal the relationships between drugs when shown by graphs. Many studies have been presented in the literature using this approach (Lim et al., 2018; Liu et al., 2016; X. Sun et al., 2019). In these studies, nodes were interpreted as drugs, and edges were interpreted as interactions between drugs.
There are also DDI studies specific to a group of drugs or to drugs used to treat a specific disease, rather than focusing on the entire drug network. For example; various studies have been conducted on cancer prevention drugs, e.g., (Jiang et al., 2020; Preuer et al., 2018) or on high blood pressure patients or medications used in the treatment, e.g., (Bacic-Vrca et al., 2010; Martha et al., 2015). In these studies, only the risks and interactions for the examined group are calculated. Therefore, in addition to a general medical corpus, additional data needs may occur with case or disease focus. However, it becomes possible to obtain faster results because it focuses on a certain region instead of the relationships formed on the whole network.
In 2018, a DDI research called Decagon was completed; it predicts the interactions of 645 drugs. (Zitnik et al., 2018). The work done by Zitnik et al. is open to development, and its models are shared so that it can be used in different projects. Accordingly, another study (ESP) (Cohen & Widdows, 2017) representing semantic predictions in pharmacological data was combined with Decagon (Burkhardt et al., 2019). The work done by Burkhardt et al. forms a resource for existing DDI researches, thanks to its rapid training and ease of reuse. In this study, in addition to the biomedical data and graph structure presented by Decagon, the infrastructure of the study conducted by Burkhardt is also used.
A. Datasets
This study uses data from the study shared by Burkhardt et al. and pre-trained vectors suitable for reuse (Burkhardt et al., 2019). The dataset cluster contains the following data from Decagon.
- 964 different polypharmacy side effects derived from a wider side effect dataset (Tatonetti et , 2012), each seen at least 500 times.
- Graph network consisting of 645 drugs and 19085 protein nodes (4,651,131 drug-drug, 18596 drug-protein node)
- Graph network hosting protein-protein and drug-protein relationships (total 8,083,300 pieces)
The LitCovid dataset was used to determine the drugs used in Covid-19 treatment. Covid-19 focused articles published in Pubmed are updated daily and added to this dataset. As of the day of the study in Mid June 2020, there are bibliographic format and summaries of about 17288 studies within the dataset.
B. DDI with Graph Convolutional Networks (GCN)
Convolutional Neural Networks (CNN) and GCN are quite similar in architectural structure. However, GCN uses graphs as input (Bastings et al., 2017). A standard GCN architecture is shown in Figure 1.
It is aimed to explore graph properties and signals for GCN (Kipf & Welling, 2019). It is assumed that each node has the properties it has from neighboring nodes and the relationships it establishes with these nodes (Z. Wu et al., 2019). Due to the Convolution layers and activation function (such as ReLU), the properties of all nodes are scanned. Depending on the study, the GCN output can be produced in different formats as a graph, featuring or representing bilateral relations.
Detecting relationships over biomedical data is one of the main study areas of GCN (Zhang et al., 2019). DDI studies or graph-based studies for the detection of side effects and ADRs are available in the literature. Through these studies, DDI predictions are carried out with different approaches. As shown in Figure 2, a graph containing drugs and proteins as nodes was created in the Decagon study.
It can be easily realized from Figure 2 that estimates of possible side effects are produced by examining drug-drug, drug-protein and protein-protein interactions. In this study, the graph shown in Figure 2 is used as the GCN input and the results of drug interaction (in the form of drug1, drug2, side_effects) are obtained as output. Using this infrastructure, Burkhardt et al. (Burkhardt et al., 2019) have also classified side effects according to diseases or organ systems.
C. Choosing the Target Drugs
In this study, a combination of three different sources was used to select the drugs whose interactionس with other drugs would be calculated. The first is the Decagon study, from which we use the network infrastructure. Another source is the LitCovid dataset, which compiles Covid-19 oriented studies on PubMed. At the last stage, it is an online system called “covid19-druginteractions.org” that predicts drug interactions and is shared by Liverpool University.
Firstly, 645 drugs from the Decagon project were chosen in the 'drug_names' dataset. In the next step, the frequency of being in the LitCovid dataset was measured for each of these drugs. The target is to identify the most mentioned drugs in Covid 19 studies published in PubMed. In order to count the number of times the drugs were mentioned, a series of operations were performed in the LitCovid data set. To avoid producing misleading results in the frequency calculation of the terms that are mentioned in the same article, only the abstract sections of the studies were searched. This way, it is easier to get more consistent results regarding the number of papers that include a drug. Data preprocessing steps have been applied to reduce the dataset only to the 'abstract' sections and make it searchable. On the normalized data, 645 drugs were subject to frequency measurement.
After frequency measurement on the LitCovid dataset, it was realized that 543 drugs were never mentioned in the dataset. The mentioning frequency of the remaining 102 drugs in the dataset was calculated as 9.06 on average. However, when we removed the 8 most frequently used drugs from the produced list, the mentioning frequency of the remaining 94 drugs drops below 3. For this reason, experiments continued by focusing only on the first 8 drugs listed in Table 1 with their mentioning frequency in the dataset. Table 1 also shows whether the drugs obtained by the LitCovid scan are available in the online system accessible from the link: www.covid19-druginteractions.org.
It has been observed that 6 of the selected drugs are also available on the 'www.covid19- druginteractions.org' website designed to show drug interactions for Covid-19.
Although Heparin and Clozapine drugs are not found in this system, they have been addressed in many studies for Covid-19 treatment (Leung et al., 2020; Perna et al., 2020; Tang et al., 2020; Thachil, 2020). Some of the studies on these drugs have been carried out in the form of clinical trials directly on patients with Covid-19. Considering their frequent history in the literature, the two drugs Heparin and Clozapine were included in this study.
D. Experiments
Experiments were conducted using a modified version of the project shared by Burkhardt, and using the pre-trained vectors of this project (GitHub - Hannahburkhardt/Predicting_ddis_with_esp: Predicting Adverse Drug-Drug Interactions with Neural Embedding of Semantic Predications, n.d.). Thus, there was no need for the re-vectorization and training of the Decagon network of medicines and proteins. Only the drugs selected in the previous section were sent as input to the side effect estimation module used in the project. This module was then updated based on the format of each input; interactions with all drugs in the network are monitored and the 5 highest scoring results are produced. It is also possible to measure whether the side effects defined in the project are included in any disease or organ system. In another update, all the side effect results resulting from the interaction of the target drugs with other drugs were classified, and the organ systems and the disease groups that these drugs may cause the most were measured. The way the project works and its differences from previous works is depicted in Figure 3.
The first part, shown as Section I in Figure 3, reflects the arthitecture of the past study, while the second part represent the arthitecture of the current study. The vector representation shown in Figure 3 was created by adding the ESP study to the Decagon study. These vectors are between 0 and 1, indicating that the effect will increase as the value gets closer to 1 and will decrease when it is closer to 0. Throughout the study, all other drugs shown in Table 1 were subject to the same steps, with Chloroquine shown in the example.
In this section, the results for each of the drugs are examined under sub-headings. In order to compare the effects of the drugs in general, the chart summarizing the side effects is shown in Figure 4; it was produced using the 645 drugs in the dataset. After the interactions of each drug were examined, the resulting side effects were classified. A drug which causes the same side effect as 25 or more drugs will have its side effects are added to the chart. This way, the most common side effects of the 8 drugs used in the experiments have been derived and visualized.
While counting side effects, the rates of the side effects were not taken into consideration. Regardless of these rates, the focus is on which disease group is mostly produced as a result of side effects. However, more detailed data by considering the rates of the side effects are reported in tables given in the sequel.
Figure 4 shares most frequently seen drugs or system side effects. Examining the chart, it is possible to infer that Heparin has more cardiovascular system side effects than other drugs, and Chloroquine causes more side effects on the immune system. These are important to create a general drug interaction projection. The reactions of the drugs will be examined in more detail in the following sub-headings.
According to Figure 4, it can be observed that hematopoietic and cardiovascular system diseases are the most common side effects of drugs. For this reason, patients with a diagnosis of Covid 19 positive and having any of these disease groups will need special attention during the treatment. Following this summary, the rest of this section addresses the following questions for each drug.
- Which are the most dangerous drugs to use in combination with a given drug?
- What are the most common side effects related to the use of other drugs with a given drug?
- Which diseases or organ systems are mostly associated with common side effects?
Ribavirin
According to the results obtained from the experiments, the drugs which have the highest rate of side effects with Ribavirin are shown in Table 2. The table shows between 2 and 5 side effects for each drug. This variability is because different edges on the graph can show the same side effects.
From the experiments, the riskiest drugs to be used with Ribavirin are listed in Table 2. Among the 644 other drugs used in the study with Ribavirin, the most common side effects are shown in Table 3. There is an important issue to be considered when examining the data in this table.
In this study, for the interaction of two drugs, a side effect prediction is always produced. These estimates vary between 0 and 1 with a value closer to 1 increases the likelihood of side effects. All the side effects produced in this study were included in the calculation. For this reason, an alternative display which considers the rates of effects is presented in the second column. In the second part of Table 3, while preparing the side effects column with the highest impact rate, the vectors created for that side effect were averaged. Other tables related to the remaining drugs have been created following the same format.
According to Table 3, due to the use of different drugs together with Ribavirin, it is likely that thrombocytopenia and dizziness complaints may be seen. Even though these are not very common in some drug combinations (such as ribavirin-pregabalin), the loss of balance and the possibility of double vision have been calculated to be quite high.
Chloroquine
The drugs with the highest rate of side effects in use with Chloroquine are listed in Table 4.
