Drug Target Commons: A Community Effort to Build a Consensus Knowledge Base for Drug-Target Interactions

Tang, Jing; Tanoli, Ziaurrehman; Ravikumar, Balaguru; Alam, Zaid; Rebane, Anni; Vähä‐Koskela, Markus; Gopalacharyulu, Peddinti; Adrichem, Arjan J. van; Wakkinen, Janica; Jaiswal, Alok; Karjalainen, Ella; Gautam, Prson; He, Liye; Parri, Elina; Khan, Suleiman A.; Gupta, Abhishekh; Ali, Mehreen; Yetukuri, Laxman; Gustavsson, Anna‐Lena; Seashore‐Ludlow, Brinton; Hersey, Anne; Leach, Andrew R.; Overington, John P.; Repasky, Gretchen A.; Wennerberg, Krister; Aittokallio, Tero

doi:10.1016/j.chembiol.2017.11.009

Cited by 154 publications

(131 citation statements)

References 25 publications

Supporting

Mentioning

128

Contrasting

Order By: Relevance

“…For that reason, it is important to develop even better sets of well-profiled inhibitors with broader target ranges and higher level of selectivity and it is promising that such efforts are moving ahead (Drewry et al, 2017;Muller et al, 2018). We are also putting our efforts in compilation of a high-quality consistent database for drug target interaction which we call Drug Target Commons (https:// drugtargetcommons.fimm.fi) (Tang et al, 2018), this kind of extensive dataset will definitely serve better for target deconvolution approaches. With further optimization of compound collections with greater diversity in terms of target spectra and chemical backbones, the method can be further explored to evaluate or predict the co-target dependencies and drug combinations effective against cancer cells.…”

Section: Discussionmentioning

confidence: 99%

Phenotypic Screening Combined with Machine Learning for Efficient Identification of Breast Cancer-Selective Therapeutic Targets

Gautam

Jaiswal

Aittokallio

et al. 2019

Cell Chemical Biology

Self Cite

View full text Add to dashboard Cite

show abstract

Section: Discussionmentioning

confidence: 99%

Phenotypic Screening Combined with Machine Learning for Efficient Identification of Breast Cancer-Selective Therapeutic Targets

Gautam

Jaiswal

Aittokallio

et al. 2019

Cell Chemical Biology

Self Cite

View full text Add to dashboard Cite

show abstract

“…AUC is used to assess classifier performances while R 2 is used for regression models. We utilized two widelyused datasets in drug-target bioactivity research: the Metz dataset (Metz et al 2011) and a subset of the Drug Target Commons (DTC) dataset (Tang et al 2018), denoted here as D4 and D5, respectively. D4 and D5 consist of 107,791 and 26,634 data points measured for the bioactivity of 1,497 drugs with 172 targets and 4,210 drugs with 599 targets, respectively.…”

Section: Supplementary Methodsmentioning

confidence: 99%

Systematic auditing is essential to debiasing machine learning in biology

Eid

Elmarakeby

Chan

et al. 2020

Preprint

View full text Add to dashboard Cite

Representational biases that are common in biological data can inflate prediction performance and confound our understanding of how and what machine learning (ML) models learn from large complicated datasets. However, auditing for these biases is not a common practice in ML in the life sciences. Here, we devise a systematic auditing framework and harness it to audit three different ML applications of significant therapeutic interest: prediction frameworks of protein-protein interactions, drug-target bioactivity, and MHC-peptide binding. Through this, we identify unrecognized biases that hinder the ML process and result in low model generalizability.Ultimately, we show that, when there is insufficient signal in the training data, ML models are likely to learn primarily from representational biases.Datasets in the life sciences have grown increasingly large and complicated. With the advent of single cell studies and biobanks, scientists are turning to machine learning (ML) to derive meaningful interpretations of these massive genomic, transcriptomic, proteomic, phenotypic, and clinical datasets. One major obstacle to the development of reliable and generalizable ML models is that auditing for biases has not been established as a common practice in life sciences ML; despite a large body of work in non-biological ML which addresses the identification and removal of algorithm biases (Zou and Schiebinger 2018). Yet, it is well known that biological datasets often suffer from representational biases stemming from evolutionary, inherent, and experimental artifacts. When these biases are not identified and eliminated, the ML process can be misled such that the model learns predominantly from the biases unique to the training dataset and, hence, is not generalizable across different datasets. In this scenario, prediction performance is inflated for the test set, but drops drastically for external predictions.Here, we demonstrate that several prominent protein-protein interaction (PPI) predictors are overwhelmingly reliant on a bias in their training dataset -to the extent that the PPI predictions become randomized (i.e., the model has no predictive power) once the bias is removed. For this reason, when applying ML to biological datasets, it is crucial to systematically audit for biases inherent in the data. This will help us to understand how and what the model is learning in order to ensure that its predictions are based on true biological insights from the data.We devised a systematic auditing framework for paired-input biological ML applications (Fig. 1a), which are widely harnessed to predict the biological relationships between two entities, e.g., physical interactions between proteins, bioactivity of drugs and their targets, or binding of MHC molecules to their antigens. We used this framework to identify biases that have confounded, over the past two decades, the ML process in three applications of great interest to the life sciences and biotech communities: protein-protein interactions, drug-target bioactivity, a...

show abstract

“…Therefore, we expected that more compounds may be annotated similarly by searching the literature which has yet been curated. A more systematic annotation may be achieved via the DrugTargetCommons platform (https://drugtargetcommons.fimm.fi/), where the crowdsourcing efforts are utilized for extracting quantitative bioactivity values of drug-target interactions from the literature (34). For the 93 cancer cell lines, their annotations have been obtained from the Cellosaurus database (35) to determine their tissues of origin.…”

Section: Annotations Of Drugs and Cell Linesmentioning

confidence: 99%

DrugComb - an integrative cancer drug combination data portal

Zagidullin

Aldahdooh

Zheng

et al. 2019

Preprint

Self Cite

View full text Add to dashboard Cite

Drug combination therapy has the potential to enhance efficacy, reduce dose-dependent toxicity and prevent the emergence of drug resistance. However, discovery of synergistic and effective drug combinations has been a laborious and often serendipitous process. In recent years, identification of combination therapies has been accelerated due to the advances in high-throughput drug screening, but informatics approaches for systems-level data management and analysis are needed. To contribute toward this goal, we created an open-access data portal (https://drugcomb.fimm.fi) where the results of drug combination screening studies are accumulated, standardized and harmonized.Through the data portal, we provided the web server to analyze and visualize users' own drug combination screening data. The users have an option to upload their data to DrugComb, as part of a crowdsourcing data curation effort. To initiate the data repository, we collected 437,932 drug combinations tested on a variety of cancer cell lines. We showed that linear regression approaches, when considering chemical fingerprints as predictors, have the potential to achieve high accuracy of predicting the sensitivity and synergy of drug combinations. All the data and informatics tools are freely available in DrugComb to enable a more efficient utilization of data resources for future drug combination discovery.

show abstract

Drug Target Commons: A Community Effort to Build a Consensus Knowledge Base for Drug-Target Interactions

Cited by 154 publications

References 25 publications

Phenotypic Screening Combined with Machine Learning for Efficient Identification of Breast Cancer-Selective Therapeutic Targets

Phenotypic Screening Combined with Machine Learning for Efficient Identification of Breast Cancer-Selective Therapeutic Targets

Systematic auditing is essential to debiasing machine learning in biology

DrugComb - an integrative cancer drug combination data portal

Contact Info

Product

Resources

About