Absence of E2f1 Negates Pro-osteogenic Impacts of p21 Absence.

Loss of p21 leads to increased bone formation post-injury; however, the mechanism(s) by which this occurs remains undetermined. E2f1 is downstream of p21 and as a transcription factor can act directly on gene expression; yet it is unknown if E2f1 plays a role in the osteogenic effects observed when p21 is differentially regulated. In this study we aimed to investigate the interplay between p21 and E2f1 and determine if the pro-regenerative osteogenic effects observed with the loss of p21 are E2f1 dependent. To accomplish this, we employed knockout p21 and E2f1 mice and additionally generated a p21/E2f1 double knockout. These mice underwent burr-hole injuries to their proximal tibiae and healing was assessed over 7 days via microCT imaging. We found that p21 and E2f1 play distinct roles in bone regeneration where the loss of p21 increased trabecular bone formation and loss of E2f1 increased cortical bone formation, yet loss of E2f1 led to poorer bone repair overall. Furthermore, when E2f1 was absent, either individually or simultaneously with p21, there was a dramatic decrease of the number of osteoblasts, osteoclasts, and chondrocytes at the site of injury compared to p21-/- and C57BL/6 mice. Together, these results suggest that E2f1 regulates the cell populations required for bone repair and has a distinct role in bone formation/repair compared to p21-/-E2f1-/-. These results highlight the possibility of cell cycle and/or p21/E2f1 being potential druggable targets that could be leveraged in clinical therapies to improve bone healing in pathologies such as osteoporosis.

XA4C: eXplainable representation learning via Autoencoders revealing Critical genes.

Machine Learning models have been frequently used in transcriptome analyses. Particularly, Representation Learning (RL), e.g., autoencoders, are effective in learning critical representations in noisy data. However, learned representations, e.g., the "latent variables" in an autoencoder, are difficult to interpret, not to mention prioritizing essential genes for functional follow-up. In contrast, in traditional analyses, one may identify important genes such as Differentially Expressed (DiffEx), Differentially Co-Expressed (DiffCoEx), and Hub genes. Intuitively, the complex gene-gene interactions may be beyond the capture of marginal effects (DiffEx) or correlations (DiffCoEx and Hub), indicating the need of powerful RL models. However, the lack of interpretability and individual target genes is an obstacle for RL's broad use in practice. To facilitate interpretable analysis and gene-identification using RL, we propose "Critical genes", defined as genes that contribute highly to learned representations (e.g., latent variables in an autoencoder). As a proof-of-concept, supported by eXplainable Artificial Intelligence (XAI), we implemented eXplainable Autoencoder for Critical genes (XA4C) that quantifies each gene's contribution to latent variables, based on which Critical genes are prioritized. Applying XA4C to gene expression data in six cancers showed that Critical genes capture essential pathways underlying cancers. Remarkably, Critical genes has little overlap with Hub or DiffEx genes, however, has a higher enrichment in a comprehensive disease gene database (DisGeNET) and a cancer-specific database (COSMIC), evidencing its potential to disclose massive unknown biology. As an example, we discovered five Critical genes sitting in the center of Lysine degradation (hsa00310) pathway, displaying distinct interaction patterns in tumor and normal tissues. In conclusion, XA4C facilitates explainable analysis using RL and Critical genes discovered by explainable RL empowers the study of complex interactions.