Category: machine learning

Artificial intelligence methods are inherently statistical in nature

April 14, 2024

Since 2018, there has been accelerating growth of applications of all kinds of artificial intelligence methods in various branches of scientific research. This has led to the use of data driven approaches in research. But, why is there a need for large amount of data to train artificial intelligence (AI) models? The answer is that … More Artificial intelligence methods are inherently statistical in nature

Enzyme activity prediction still revolves around the enzyme activity-protein structure correlate

March 29, 2024

From the theoretical perspective, enzyme structure is perhaps the easiest correlate for enzyme activity levels given the intuitive nature of structure-function correlates. But, the challenge lies in extracting useful features from a structural depiction that could be correlated with enzyme activity. To this end, the field of enzyme activity prediction still awaits comprehensive proteome level … More Enzyme activity prediction still revolves around the enzyme activity-protein structure correlate

Many manufacturing plants in the pharmaceutical industry are moving into the big data era

March 15, 2024

Big data has been touted to revolutionize many industries in the near future. Indeed, the era of big data has arrived in many manufacturing industries such as the petrochemical and biopharmaceutical industries. In the case of manufacturing plants of pharmaceutical industries, there is a movement towards digitisation of all manufacturing operations to help capture all … More Many manufacturing plants in the pharmaceutical industry are moving into the big data era

Machine learning models are derived based on correlations rather than mechanistic insights

February 10, 2024

Machine learning tools have found utility in a variety of fields ranging from material synthesis to bioinformatics and medicine. However, little is known about how machine learning models are derived. Basically, machine learning models are obtained based on the work of pattern recognition algorithms gleaning unseen correlations between variables in a large dataset, the latter … More Machine learning models are derived based on correlations rather than mechanistic insights

Active learning is the next frontier in machine learning

January 7, 2024

Machine learning has seen a large growth in applications over the past few years, and has impacted on the fields of clinical diagnostics, security, and drug discovery. The typical workflow in machine learning involves the preprocessing of input data into standardized format readable by computers, random division of data into training and validation sets, training … More Active learning is the next frontier in machine learning

Application of machine learning tools in predicting enzyme activity

December 18, 2021

Currently, the holy grail in application of machine learning to predicting enzyme performance lies in using sequence information for understanding the relative levels of enzyme activity. Such a task transverses multiple steps of biological abstraction and is not trivial. Specifically, enzyme activity level is predicated primarily on the affinity between the enzyme for the substrate … More Application of machine learning tools in predicting enzyme activity

Supervised machine learning needs representative training set to build accurate predictive models

December 4, 2021

Machine learning algorithms are generally classified into unsupervised and supervised machine learning tools. Specifically, unsupervised machine learning algorithms are capable of discerning patterns in data without the need for prior training on similar data. On the other hand, supervised machine learning tools require the provision of training data to help build a model that forms … More Supervised machine learning needs representative training set to build accurate predictive models

Generic methods for building machine learning models may not yield the best optimized model

February 5, 2021

Current trend in machine learning splits practitioners in the field into two camps: (i) those that use established machine learning tools to build models on new data, and (ii) those that design and conceptualize new machine learning tools. Typically, many novice users of machine learning tools would iteratively test established machine learning tools such as … More Generic methods for building machine learning models may not yield the best optimized model

Predictions from biological deep learning models are hard to interpret due to their “black box” nature

November 11, 2020

Involving the use of neural networks, deep learning is increasingly applied in many facets of biological research to glean insights into complex data. Such models could proffer patterns in complex data otherwise hidden from view, but which may help inform future experimentations or seed new research inquiries. In actual implementation, deep learning approaches utilises a … More Predictions from biological deep learning models are hard to interpret due to their “black box” nature

Machine learning models are derived based on correlations rather than mechanistic insights

November 4, 2020

Machine learning tools have found utility in a variety of fields ranging from material synthesis to bioinformatics and medicine. However, little is known about how machine learning models are derived. Basically, machine learning models are obtained based on the work of pattern recognition algorithms gleaning unseen correlations between variables in a large dataset, the latter … More Machine learning models are derived based on correlations rather than mechanistic insights