Machine learning alternative to systems biology should not solely depend on data

Abstract

In recent years, artificial intelligence (AI)/machine learning has emerged as a plausible alternative to systems biology for the elucidation of biological phenomena and in attaining specified design objective in synthetic biology. Although considered highly disruptive with numerous notable successes so far, we seek to bring attention to both the fundamental and practical pitfalls of their usage, especially in illuminating emergent behaviors from chaotic or stochastic systems in biology. Without deliberating on their suitability and the required data qualities and pre-processing approaches beforehand, the research and development community could experience similar ‘AI winters’ that had plagued other fields. Instead, we anticipate the integration or combination of the two approaches, where appropriate, moving forward.

Identifying effective evolutionary strategies for uncovering reaction kinetic parameters under the effect of measurement noises

“…Furthermore, the mathematical model of the change in component levels with time based on such interactions (i.e., rate laws or reaction kinetics), can be used to flexibly simulate, and predict component dynamics under broad conditions. From an application perspective, the mechanistic model also provides a structure grounded in first principles to help reduce sample size and datapoint requirement for training, effect of biological and technical noises, data leakage, as well as batch effects [2]. The model also grounds predictions in molecular mechanism, thereby allowing for experimental verification, and enabling explainability.…”

“…In an earlier commentary, we have set out the critical role of explicitly modelling the network and reaction kinetics for effective learning about biological systems [2]. For metabolic engineering [3], synthetic biology [4], or even precision drug endeavors [5], the holistic modelling of physico-chemical interactions among molecular components (systems biology [6]) is necessary for the reconstruction and elucidation of emergent behaviors and the identification of system-wide mechanisms and effects.…”

“…While systems biology has begun its nascent transition from explanatory modelling of fitted data to the predictive modelling of unseen data [2, 8], there remains two broad objectives in the field: The learning of a mechanistic model of the system for predicting unobserved dynamics, a shift from the earlier, more restricted objective of calibrating the parameters of a given model to best reproduce the fitted data [9]; The finding of an optimal set of system input variables, given the design objective(s) for the biological system. This is typically done by repeatedly simulating the system, according to values sampled from the input variable hyperspace.…”

“…While systems biology has begun its nascent transition from explanatory modelling of fitted data to the predictive modelling of unseen data [2,8], there remains two broad objectives in the field:…”

Desenvolvimento e validação de um índice de multimorbidade para idosos e seu uso no prognóstico de declínio funcional em três anos

“…Uma última observação a fazer é que, embora os novos métodos de análise de dados conhecidos por aprendizado de máquina sejam muito promissores e que eles possam trazer grandes desenvolvimentos para a ciência biomédica, esses métodos são muito sensíveis a artefatos relacionados a questões metodológicas dos estudos e a ruídos que não deveriam ser incorporados aos modelos (80).…”