Rakesh Sharma | Search bioRxiv

bioRxiv2020

The spatial proteome of the human intervertebral disc reveals architectural changes in health, ageing and degeneration

The spatiotemporal proteome of the intervertebral disc (IVD) underpins its integrity and function. We present DIPPER, a deep and comprehensive IVD proteomic resource comprising 94 genomewide profiles from 17 individuals. To begin with, protein modules defining key directional trends spanning the lateral and anteroposterior axes were derived from high-resolution spatial proteomes of intact young cadaveric lumbar IVDs. They revealed novel region-specific profiles of regulatory activities, and displayed potential paths of deconstruction in the level- and location-matched aged cadaveric discs. Machine learning methods predicted a "hydration matrisome" that connects extracellular matrix with MRI intensity. Importantly, the static proteome used as point-references can be integrated with dynamic proteome (SILAC/degradome) and transcriptome data from multiple clinical samples, enhancing robustness and clinical relevance. The data, findings and methodology, available on a web interface, will be valuable references in the field of IVD biology and proteomic analytics.

systems biology↗

bioRxiv2020

Crowdsourcing Machine Intelligence Solutions to Accelerate Biomedical Science: Lessons learned from a machine intelligence ideation contest to improve the prediction of 3D domain swapping

Machine intelligence methods, including natural language processing, computer vision, machine vision, artificial intelligence, and deep learning approaches, are rapidly evolving and play an essential role in biomedicine. Machine intelligence methods could help to accelerate image analyses aid in building complex models capable of interpretation beyond cognitive limitations and statistical assumptions in biomedicine. However, irrespective of the democratization via accessible computing and software modules, machine intelligence handiness is scarce in the setting of a traditional biomedical research laboratory. In such a context, collaborations with bioinformatics and computational biologists may help. Further, the biomedical diaspora could also seek help from the expert communities using a crowdsourcing website that hosts machine intelligence competitions. Machine intelligence competitions offer a vast pool of seasoned data scientists and machine intelligence experts to develop solutions through competition portals. An alternate approach to improve the adoption of machine intelligence in biomedicine is to offer machine intelligence competitions as part of scientific meetings. In this paper, we discuss a structured methodology employed to develop the machine intelligence competition as part of an international bioinformatics conference. The competition leads to developing a novel method through crowdsourcing to solve a challenging problem in biomedicine - predicting probabilities of proteins that undergo 3D domain swapping. As a biomedical science conference focused on computational methods, the competition received multiple entries that ultimately helped improve the predictive modeling of 3D domain swapping using sequence information.

bioinformatics↗