Receive a weekly summary and discussion of the top papers of the week by leading researchers in the field.

General General

UniPath: a uniform approach for pathway and gene-set based analysis of heterogeneity in single-cell epigenome and transcriptome profiles.

In Nucleic acids research ; h5-index 217.0

Recent advances in single-cell open-chromatin and transcriptome profiling have created a challenge of exploring novel applications with a meaningful transformation of read-counts, which often have high variability in noise and drop-out among cells. Here, we introduce UniPath, for representing single-cells using pathway and gene-set enrichment scores by a transformation of their open-chromatin or gene-expression profiles. The robust statistical approach of UniPath provides high accuracy, consistency and scalability in estimating gene-set enrichment scores for every cell. Its framework provides an easy solution for handling variability in drop-out rate, which can sometimes create artefact due to systematic patterns. UniPath provides an alternative approach of dimension reduction of single-cell open-chromatin profiles. UniPath's approach of predicting temporal-order of single-cells using their pathway enrichment scores enables suppression of covariates to achieve correct order of cells. Analysis of mouse cell atlas using our approach yielded surprising, albeit biologically-meaningful co-clustering of cell-types from distant organs. By enabling an unconventional method of exploiting pathway co-occurrence to compare two groups of cells, our approach also proves to be useful in inferring context-specific regulations in cancer cells. Available at

Chawla Smriti, Samydurai Sudhagar, Kong Say Li, Wang Zhenxun, Tam Wai Leong, Sengupta Debarka, Kumar Vibhor


General General

Seizure Detection and Epileptogenic Zone Localisation on Heavily Skewed MEG Data using RUSBoost Machine Learning Technique.

In The International journal of neuroscience

Epilepsy is a neurological disorder which is characterised by recurrent and involuntary seizures. Magnetoencephalography (MEG) is clinically used as a presurgical tool in locating the epileptogenic zone by localising either interictal epileptic discharges (IEDs) or ictal activities. The localisation of ictal onset provides reliable and more accurate seizure onset zones rather than localising the IEDs. Ictals or seizures are presently detected during MEG analysis by manually inspecting the recorded data. This is laborious when the duration of recordings is longer. Hence, we propose a method which uses statistical features such as short-time permutation entropy (STPE), gradient of STPE (GSTPE), short-time energy (STE) and short-time mean (STM) extracted from the ictal and interictal MEG data of drug resistant epilepsy patients group. Since the data is heavily skewed, RUSBoost machine learning technique is used to classify the data into ictal and inter-ictal by using the four feature vectors. This method is further used for localizing the epileptogenic region using region-specific classifications by means of RUSBoost algorithm. The accuracy obtained for seizure detection is 93.4% with an area-under-curve (AUC) of 0.97. The specificity and sensitivity for the same are 93%. The localization accuracies for each region are in the range of 88.1% - 99.1%.

Bhanot Nipun, Mariyappa N, Anitha H, Bhargava G K, Velmurugan J, Sinha S


EEG, MEG, RUSBoost, artificial neural networks, classification, epileptogenic zone, ictal, inter-ictal, machine learning, permutation entropy, seizure detection

General General

StackNet-DenVIS: a multi-layer perceptron stacked ensembling approach for COVID-19 detection using X-ray images.

In Physical and engineering sciences in medicine

The highly contagious nature of Coronavirus disease 2019 (Covid-19) resulted in a global pandemic. Due to the relatively slow and taxing nature of conventional testing for Covid-19, a faster method needs to be in place. The current researches have suggested that visible irregularities found in the chest X-ray of Covid-19 positive patients are indicative of the presence of the disease. Hence, Deep Learning and Image Classification techniques can be employed to learn from these irregularities, and classify accordingly with high accuracy. This research presents an approach to create a classifier model named StackNet-DenVIS which is designed to act as a screening process before conducting the existing swab tests. Using a novel approach, which incorporates Transfer Learning and Stacked Generalization, the model aims to lower the False Negative rate of classification compensating for the 30% False Negative rate of the swab tests. A dataset gathered from multiple reliable sources consisting of 9953 Chest X-rays (868 Covid and 9085 Non-Covid) was used. Also, this research demonstrates handling data imbalance using various techniques involving Generative Adversarial Networks and sampling techniques. The accuracy, sensitivity, and specificity obtained on our proposed model were 95.07%, 99.40% and 94.61% respectively. To the best of our knowledge, the combination of accuracy and false negative rate obtained by this paper outperforms the current implementations. We must also highlight that our proposed architecture also considers other types of viral pneumonia. Given the unprecedented sensitivity of our model we are optimistic it contributes to a better Covid-19 detection.

Autee Pratik, Bagwe Sagar, Shah Vimal, Srivastava Kriti


Covid-19, Deep neural networks, Generative adversarial networks, Image segmentation, Stacked generalization, Transfer learning

General General

DeepPurpose: A Deep Learning Library for Drug-Target Interaction Prediction.

In Bioinformatics (Oxford, England)

SUMMARY : Accurate prediction of drug-target interactions (DTI) is crucial for drug discovery. Recently, deep learning (DL) models for show promising performance for DTI prediction. However, these models can be difficult to use for both computer scientists entering the biomedical field and bioinformaticians with limited DL experience. We present DeepPurpose, a comprehensive and easy-to-use deep learning library for DTI prediction. DeepPurpose supports training of customized DTI prediction models by implementing 15 compound and protein encoders and over 50 neural architectures, along with providing many other useful features. We demonstrate state-of-the-art performance of DeepPurpose on several benchmark datasets.


SUPPLEMENTARY INFORMATION : Supplementary data are available at Bioinformatics online.

Huang Kexin, Fu Tianfan, Glass Lucas M, Zitnik Marinka, Xiao Cao, Sun Jimeng


Ophthalmology Ophthalmology

Evaluation of Four Artificial Intelligence-Assisted Self-Diagnosis Apps on Three Diagnoses: Two-Year Follow-Up Study.

In Journal of medical Internet research ; h5-index 88.0

BACKGROUND : Consumer-oriented mobile self-diagnosis apps have been developed using undisclosed algorithms, presumably based on machine learning and other artificial intelligence (AI) technologies. The US Food and Drug Administration now discerns apps with learning AI algorithms from those with stable ones and treats the former as medical devices. To the author's knowledge, no self-diagnosis app testing has been performed in the field of ophthalmology so far.

OBJECTIVE : The objective of this study was to test apps that were previously mentioned in the scientific literature on a set of diagnoses in a deliberate time interval, comparing the results and looking for differences that hint at "nonlocked" learning algorithms.

METHODS : Four apps from the literature were chosen (Ada, Babylon, Buoy, and Your.MD). A set of three ophthalmology diagnoses (glaucoma, retinal tear, dry eye syndrome) representing three levels of urgency was used to simultaneously test the apps' diagnostic efficiency and treatment recommendations in this specialty. Two years was the chosen time interval between the tests (2018 and 2020). Scores were awarded by one evaluating physician using a defined scheme.

RESULTS : Two apps (Ada and Your.MD) received significantly higher scores than the other two. All apps either worsened in their results between 2018 and 2020 or remained unchanged at a low level. The variation in the results over time indicates "nonlocked" learning algorithms using AI technologies. None of the apps provided correct diagnoses and treatment recommendations for all three diagnoses in 2020. Two apps (Babylon and Your.MD) asked significantly fewer questions than the other two (P<.001).

CONCLUSIONS : "Nonlocked" algorithms are used by self-diagnosis apps. The diagnostic efficiency of the tested apps seems to worsen over time, with some apps being more capable than others. Systematic studies on a wider scale are necessary for health care providers and patients to correctly assess the safety and efficacy of such apps and for correct classification by health care regulating authorities.

Ćirković Aleksandar


artificial intelligence, mHealth, machine learning, medical diagnosis, mobile apps

General General

An Artificial Intelligence-Based, Personalized Smartphone App to Improve Childhood Immunization Coverage and Timelines Among Children in Pakistan: Protocol for a Randomized Controlled Trial.

In JMIR research protocols ; h5-index 26.0

BACKGROUND : The immunization uptake rates in Pakistan are much lower than desired. Major reasons include lack of awareness, parental forgetfulness regarding schedules, and misinformation regarding vaccines. In light of the COVID-19 pandemic and distancing measures, routine childhood immunization (RCI) coverage has been adversely affected, as caregivers avoid tertiary care hospitals or primary health centers. Innovative and cost-effective measures must be taken to understand and deal with the issue of low immunization rates. However, only a few smartphone-based interventions have been carried out in low- and middle-income countries (LMICs) to improve RCI.

OBJECTIVE : The primary objectives of this study are to evaluate whether a personalized mobile app can improve children's on-time visits at 10 and 14 weeks of age for RCI as compared with standard care and to determine whether an artificial intelligence model can be incorporated into the app. Secondary objectives are to determine the perceptions and attitudes of caregivers regarding childhood vaccinations and to understand the factors that might influence the effect of a mobile phone-based app on vaccination improvement.

METHODS : A mixed methods randomized controlled trial was designed with intervention and control arms. The study will be conducted at the Aga Khan University Hospital vaccination center. Caregivers of newborns or infants visiting the center for their children's 6-week vaccination will be recruited. The intervention arm will have access to a smartphone app with text, voice, video, and pictorial messages regarding RCI. This app will be developed based on the findings of the pretrial qualitative component of the study, in addition to no-show study findings, which will explore caregivers' perceptions about RCI and a mobile phone-based app in improving RCI coverage.

RESULTS : Pretrial qualitative in-depth interviews were conducted in February 2020. Enrollment of study participants for the randomized controlled trial is in process. Study exit interviews will be conducted at the 14-week immunization visits, provided the caregivers visit the immunization facility at that time, or over the phone when the children are 18 weeks of age.

CONCLUSIONS : This study will generate useful insights into the feasibility, acceptability, and usability of an Android-based smartphone app for improving RCI in Pakistan and in LMICs.



Kazi Abdul Momin, Qazi Saad Ahmed, Khawaja Sadori, Ahsan Nazia, Ahmed Rao Moueed, Sameen Fareeha, Khan Mughal Muhammad Ayub, Saqib Muhammad, Ali Sikander, Kaleemuddin Hussain, Rauf Yasir, Raza Mehreen, Jamal Saima, Abbasi Munir, Stergioulas Lampros K


AI, EPI, LMICs, Pakistan, artificial intelligence, mHealth, personalized messages, routine childhood immunization, routine immunization, smartphone apps, vaccine-preventable illnesses