Knowledge-guided artificial intelligence technologies for decoding complex multiomics interactions in cells

Abstract

Cells survive and proliferate through complex interactions among diverse molecules across multi-omics layers. Conventional experimental approaches for identifying these interactions have built a firm foundation for molecular biology, but their scalability is gradually becoming inadequate compared to the rapid accumulation of multi-omics data measured by highthroughput technologies. Therefore, the need for data-driven computational modeling of interactions within cells has been highlighted in recent years. The complexity of multi-omics interactions is primarily due to their nonlinearity. That is, their accurate modeling requires intricate conditional dependencies, synergies, or antagonisms between considered genes or proteins, which retard experimental validations. Artificial intelligence (AI) technologies, including deep learning models, are optimal choices for handling complex nonlinear relationships between features that are scalable and produce large amounts of data. Thus, they have great potential for modeling multi-omics interactions. Although there exist many AIdriven models for computational biology applications, relatively few explicitly incorporate the prior knowledge within model architectures or training procedures. Such guidance of models by domain knowledge will greatly reduce the amount of data needed to train models and constrain their vast expressive powers to focus on the biologically relevant space. Therefore, it can enhance a model's interpretability, reduce spurious interactions, and prove its validity and utility. Thus, to facilitate further development of knowledge-guided AI technologies for the modeling of multi-omics interactions, here we review representative bioinformatics applications of deep learning models for multi-omics interactions developed to date by categorizing them by guidance mode.

Missing data in multi-omics integration: Recent advances through artificial intelligence

“…These biologically informed constraints are likely critical to develop successful approaches, as relying on over-parametrized models such as DNN's to learn non-spurious, inspectable relationships is not a reliable strategy (Lee and Kim, 2022 ). However, the strength of over-parametrized models in learning complex relationships cannot be ignored.…”

“…If the generation of these latent factors were instead guided by domain knowledge, resulting groupings may be more likely to hold relevance and rely less on post-hoc analyses for contextualization within the scientific domain. Thus, recognizing the need for greater interpretability of the more opaque integration methods, efforts to combine biological-knowledge-injecting methods such as graph representations or other hard constraints on parameter representations with machine learning methods are ongoing (Noor et al, 2019 ; Lee and Kim, 2022 ), and evaluation and development of mechanisms for handling missing data within these frameworks will be necessary.…”

“…Many are focused on specific application domains such as metabolic engineering (Helmy et al, 2020 ), precision medicine (Hamamoto et al, 2019 ), amongst others (e.g., Mann et al, 2021 ; Lin et al, 2022 ; Zhou et al, 2022 ) and a majority focused on cancer research (e.g., Wang and Gu, 2016 ; Biswas and Chakrabarti, 2020 ; Nicora et al, 2020 ; Cai et al, 2022 ). Finally, a handful of reviews are dedicated to discussion of subsets of general AI and ML approaches to multi-omics data integration (e.g., Li et al, 2016 ; Huang et al, 2017 ; Kim and Tagkopoulos, 2018 ; Picard et al, 2021 ; Reel et al, 2021 ; Lee and Kim, 2022 ) with some attention given to limitations of methods, such as interpretability of models.…”

Bibliometric analysis and visualisation of research hotspots and frontiers on omics in osteosarcoma

“…Omics technology is used to study a series of molecules and their interactions involved in the whole process of gene expression from a systems-level, mainly including genomics, transcriptomics, proteomics and metabolomics (Jeong et al2023). With the advances in the technologies and tools for generating and processing large omics data, and the application of arti cial intelligence methodologies for deciphering complex multi-omics interaction, omics technology becomes a powerful approach to decipher the mechanistic details of gene expression (Lee et al 2022). Omics technology is excepted to complement current clinical and pathology evaluations and guide personalized cancer management by discovering previously obscure sub-types with clinical implications and identifying patients' prognoses, which help in revealing the molecular mechanisms and heterogeneity of OS, thereby improving the prognosis of patients (Pan et al 2021).…”

Advancing pediatric health: the multifaceted scope of Clinical and Experimental Pediatrics

“…During the review process, it was identified that CEP published a paper outside its scope. 2) This review article introduced a model of deep learning at the cellular level and provided information on the concept of multiomics, an expanding research area. Topics are not limited to pediatric health or diseases but include generalized research applicable to all basic medical fields.…”