We are dedicated to understanding the genomic causal behind pediatric cancer, with combined computational and experimental approaches. The goal of our team is to translate our findings in the lab to assist patient management in clinic, including diagnosis and treatment. We hope our work could contribute to the arrival of next generation clinic management pattern in a near future. To this end, our work could be largely grouped into the following three directions.

Develop state of the art computational methods to uncover novel driver genomic aberrations

The applying of second-generation sequencing to cancer genomics had greatly advanced our understanding of the disease by identifying many driver aberrations in each individual tumor type. However, there are still a certain percentage of tumors lack a clear driver. For example, in ~20% pediatric neuroblastomas and ~7% of pediatric B-ALL, among others, we still don’t know the driver aberrations behind. This indicates a lack of power for current analytical methods to identify the variants and establish their causal relationship with disease. In collaboration with Dr. Zhang at St. Jude Children’s Research Hospital, we recently developed novel computational methods called cis-X to expand our driver discovery ability from coding exons in the genome to noncoding regions. The later consists ~98% of genome in human and are known to be enriched with functional regulatory elements. Our new approach combined allelic specific expression and outlier high expression of a given gene, which are two pivotal signatures of cis activation but not well incorporated in existing tools in this field. Importantly, cis-X was designed to analyze individual cancer genome, thus significantly increased the power of noncoding variant analysis which generally required a large cohort (>1,000) in previous studies. A pilot analysis with only 13 pediatric T-ALLs had already unveiled novel oncogenic regulatory mutations driving TAL1 and a potential oncogene PRLR which was completely unknown in pediatric T-ALL. Wilder application of cis-X could contribute more disease associated noncoding variants and genes in this field, yield not only a better understand of the disease but also potential novel actionable variants/genes for clinic consideration. Following this, we are working on further increase the power of regulatory noncoding variant analysis by integrating multiple dimensional data in the lab.

Genomic and epigenomic profiling of pediatric cancer

Equipped with the cutting-edge computational methods, we are actively studying more pediatric cancer genomes with an emphasize of pediatric patients in China. It’s easy to note that the majority of genomic analysis published so far were based on western populations. Little was known if there was a difference between western and eastern patients in terms of either mutation types or frequencies or their association with clinic behavior of patients. In collaborations with Dr. Shuhong Shen and team at Shanghai Children’s Medical Center, we are currently analyzing thousands of pediatric cancer genomes to patch the hole and bridge the lab and clinic.

Cloud computing based clinical genomics

Pediatric cancer is rare disease. To achieve the goal of translating genomic findings in the lab to bedside, a large cohort of patients under standard treatment protocol and computational analysis pipelines are needed. We are lucky to join the Chinese Children Cancer Group - Acute Lymphoblastic Leukemia (CCCG-ALL) which have 20 hospitals from 13 provinces/regions and cover ~65% of population in China. The CCCG-ALL group apply standardized diagnosis and treatment protocol among different hospitals. On top of this, we have developed a cloud-based computing pipeline and integration system to add the genomic aspect into the big picture. Launched in August 2020, our system integrates experiment monitoring, computational analysis and real time clinic report tracking. Within the first 4 months, we have recruited and analyzed ~200 ALL transcriptomes in this working group utilizing this platform. We aim to recruit ~10,000 patients under this unified clinic-computational protocol and to integrate with clinic data applying AI approaches to refine patient risk stratification in clinic.