Editorial for BFG special issue: Computational genomics for precision medicine and personalized healthcare

Given the spectacular advancement in recent years both in high-throughput genomic sequencing technology as well as performance of computational methodologies simultaneously, computational genomics has immense potential for revolutionizing personalized healthcare and precision medicine in the near future. Traditional computational modelling has focussed on maximizing population level predictive accuracy, rather than individual level precision, and out of distribution points were treated as outliers. However, for computational models to be used in high stakes healthcare applications, the model predictions must cater to each and every individual, instead of a ‘one size fits all approach’. This special issue brings together a range of high impact review articles and overview briefings on recent advancements in the field of computational genomics which makes it highly timely. The articles can be broadly divided into the following three overlapping categories, viz., methods pertaining to biological insights through technology, AI enabled personalized healthcare, and in-silico drug repurposing strategies.

Three of the works present state-of-the-art discovery focussed methodologies in computational genomics for developing the biological science of precision medicine. Wilkinson et al. [1] demonstrated that third generation genomic sequencing can be used for bacterial infection identification in a timeframe comparative to that for joint revision surgery, thereby raising the possibility of real time diagnosis of joint infections. They point out however that further work is needed to analyse whether prediction of antibiotic resistance can also be achieved in similar clinically relevant timeframes. Shorthouse et al. [2] present a comprehensive review of recent computational methods that investigate how genetic variations influence diseased cellular behaviour, especially loss of phenotypic functions due to protein misfolding, which has high significance for personalized patient treatment in light of significant advances in high-quality, low-cost high-throughput sequencing technologies. Bahar et al. [3] delve into the recently demonstrated promise of Gaussian Network Models as an effective tool for modelling chromatin structural dynamics from Hi-C data at different hierarchical levels from single gene loci to full chromosomes as a basis for insights into gene co-expression, transcriptional regulation, and epigenetic modifications. Halder et al. [4] provide an in-depth review of two protein (CTCF and Cohesin) factor-specific loop interaction from the perspective of how chromatin loops and structures can help understand gene regulation and genome organization, including recent analysis methods involving deep learning.

In contrast, two of the other works present recent advancements in AI enabled high performance computing for personalized healthcare and individualized treatments, while handling patient heterogeneity. Mukhopadhyay et al. [5] provide a comprehensive overview of how multimodal omics data (genomics, transcriptomics, proteomics, and metabolomics) can be leveraged alongside clinical data and electronic health records to provide patient stratification and precision treatments. They present in a very accessible way recent methodological progress in deep learning and network models these seemingly different data modalities into a harmonized inference space. Similarly, Saha et al. [6] survey the landscape of multi-modal AI based approaches in breast cancer detection and prognosis, particularly from the point of transparency and interpretability, for safer and trustworthy translation to practice. AI is transitioning from unimodal to multimodal currently, thus opening up a whole vista of possibilities, including in the field of computational genomics, where it can be fused with other modalities like imaging for joint decision systems.

Finally, the remaining two papers delve into prospects of in-silico drug repurposing with recent computational modelling approaches, as cheap and safe alternatives to new drug discovery. Saha et al. [7] introduce an overview of computational methods in drug repurposing for the treatment of viral diseases, using Monkeypox as a case study, though the insights are generalizable to other similar disease scenarios. They show how gene expression patterns and protein–protein interaction networks can be used for predictive modelling of hosts’ response to pathogens and drugs. Similarly, Halder et al. [8] explore how intermolecular interactions of protein–protein complexes help in in-silico drug discovery and drug repurposing by analysing behaviour of target antibodies. They present their findings in the context of Ebolavirus and Marburgvirus (common family Filoviridae), but again on a methodological scale, making these inferences potentially generalizable to other disease scenarios.

Thus this special issue showcases the cutting edge in computational genomics from the perspectives of scientific knowledge emerging from technological advancements in the field, AI enabled multimodal decision systems for precision medicine despite patient heterogeneity, as well as computational modelling for in-silico drug repurposing and personalized treatment. The editors hope that this special issue will be of timely interest to the discerning readership of the Briefings in Functional Genomics.

Funding

None declared.