The journal article “Integrating network, sequence and functionalfeatures using machine learning approaches towards identification of novelAlzheimer genes” by Salma Jamal, Sukriti Goyal, Asheesh Shanker and AbhinavGrover discussed the neurodegenerative disorder causing dementia in the elderlycommunity called Alzheimer’s disease (AD) and trying to find a better way toidentify Alzheimer genes.
Alzheimer’s is a disease that destroys memory by short term memory loss as well as loss of intellectualand social skills. It is the most common cause of dementia. There is no treatment forAlzheimer’s and cause is still relatively unknown.
There is a great need tolearn more about cause and potential treatment to slow this degeneratingdisease. It has been said that genetics plays a role. The article aims to lookat machine learning as a way find potential AD genes that my revealsusceptibility of getting the disease, which could lead finding and developing new anti-Alzheimerdrugs. To identifypotential AD associated genes, they looked at protein-protein interaction networks,sequence features and functional annotations as well they looked at therapeutictargets. The authors wrote a fine abstractthat was easy to read and understandable, it clearly described what theAlzheimer’s disease was and why there was a need to further look in to thistopic. The abstract provides a little insight to their results and concludedwith a statement about where machine learning can go in the future by statingthat computational studies like the one they completed may be able to improveour understanding of the causes of Alzheimer which may be true but machine learningneeds to be validated to say it can definitively pinpoint the cause of adisease. The introduction section provided great background onAlzheimer’s disease providing enough details to understand why they thought itwas important to look into the topic.
The provided not only enough informationon the disease but also provided details on some of the other methods that havebeen used to look at Alzheimer disease as well. The article talked about what wasalready known but where we lacked knowledge like that the Alzheimer-associatedgenes already known do not cover a substantial portion of the human genome,meaning that there are countless disease associate genes that can discovered.This is one reason why they are looking in to using machine learning.
Overall the method sections seem well written as it describedand explained the databases in which they used for annotation, visualizationand sequences. It also explained why each feature was pick. An example “Wecalculated 9 topological properties …… radiality are the indicators of thecentrality of a node in a biological network” (pg 2-3). This is good concept in writing amethods section in letting people know what it, and what would be measured. Ifelt the article does a good job of trying to sum up such a complicated methodfor those who may be unfamiliar with such databases and “computer” languageused. The authors indicated the specificstatistical procedures and formulas used for calculations. Although the methodswere described in details I don’t think others that are less familiar with thetopic could exactly repeat this study.
For people who have an extensivebackground with proteins and the databases used they would be more likely torepeat the study. When discussing the results, (resultsand discussion were together) they again provide what their exact objective waswhich was “to identify potential Alzheimer genes based on the extraction of their network, sequences andfunctional properties using machine learning approaches” (pg 5). Which was metwhen they were able to find 13 genes based on the performance from various machine learning algorithms.The algorithms were built from the model systems using training set thatclassified the disease genes from the non-disease genes. The results that they obtain seemed to make sense base ofwhat is known about the 13 genes they found.
they that considered for further analysis. The fact that theyprovided/discussed information that was already known about each of the 13genes helped support the claim that the genes that the machine learning systemidentified really could play a role in the predicted development of Alzheimer disease. They were also ableto identify a current anti-Alzheimer disease drug. The tables and graphs providedalso made it easy to look at the comparisons that were describing. The articlemakes reference to several additional files in which the results were presentbut the tables and how the results were generated were not and would have beenhelpful in understanding the whole picture. The authors also citedreference material that they have used when writing the paper citing over 100articles.
Authors also cite other paper they have contributed some of which Idid think were relevant the writing of this particular paper. Overall this articleis takes an investigative look into machine learning which within the field ofbiotechnology a growing field. The authors do a great job intergrading thecomputer science aspects of machine learning and incorporate biology in hopesvastly improved understanding of Alzheimer disease causes. The strengths ofthis article are that it uses a largedataset, with reasonable coverage across genes totaling,56405.
Also, that fact that some of the 13 genes that they identified had been knownfor regulation of synapses, activate mechanisms that are key players inneuronal apoptosis and other neurological