Nitecap: An Exploratory Circadian Analysis Web Application

Abstract

Circadian omics analyses present investigators with large amounts of data to consider and many choices for methods of analysis. Visualization is crucial as rhythmicity can take many forms and p-values offer an incomplete picture. Yet statically viewing the entirety of high-throughput datasets is impractical, and there is often limited ability to assess the impact of choices, such as significance threshold cutoffs. Nitecap provides an intuitive and unified web-based solution to these problems. Through highly responsive visualizations, Nitecap enables investigators to see dataset-wide behavior. It supports deep analyses, including comparisons of two conditions. Moreover, it focuses upon ease-of-use and enables collaboration through dataset sharing. As an application, we investigated cross talk between peripheral clocks in adipose and liver tissues and determined that adipocyte clock disruption does not substantially modulate the transcriptional rhythmicity of liver but does advance the phase of core clock gene Bmal1 (Arntl) expression in the liver. Nitecap is available at nitecap.org and is free-to-use.

Keywords: circadian analysis; principal component analysis; rhythmicity analysis; visualization; web application.

Figure 1.

Nitecap interface. The main Nitecap spreadsheet-viewing page allows investigators to quickly see profile plots and computed statistics. The displayed plot and statistics update in real time as the user scans through the list of spreadsheet rows (containing genes, transcripts, proteins, etc.), allowing investigators to obtain quick intuition about their dataset, such as the robustness of rhythmicity at various cutoffs. The spreadsheet rows can be sorted by any chosen statistic, including user-provided ones, by selecting it from the statistics panel on the right. Abbreviations: KO = knock-out; WT = wild-type.

Figure 2.

Nitecap implementation. Typical workflow in Nitecap. First, the front end sends the request to run an algorithm (1). This request passes through the load balancer and is received by the Nitecap server (2) which is deployed as a containerized service in Amazon ECS. The server puts the data needed for the algorithm in the S3 bucket (3) and instructs the orchestrator to run the desired algorithm (4). The orchestrator then starts the appropriate Lambda function (5) where the algorithm is run. During the run, the status notifications are sent to the API Gateway (6) which pushes them to the front end (7)via the WebSocket protocol. The Amazon Dynamo database is used to keep track of which users are actively connected to Nitecap via the WebSocket protocol. Abbreviations: API = application programming interface; ECS = Elastic Container Service.

Figure 3.

PCA comparison—DRF vs NRF. RNA-seq from mouse liver tissue under DRF and NRF. PCA of the samples using (a) the top 500 rhythmic genes in NRF, (b) the top 500 rhythmic genes in DRF, or (c) all genes. In all cases, expression values were log-transformed and z scored prior to taking the PCA. Rhythmicity testing by JTK_CYCLE. Abbreviations: PCA = principal component analysis; DRF = day-restricted feeding; NRF = night-restricted feeding.

Figure 4.

PCA comparison—Bmal1 KO. (a-b) RNA-seq was performed on liver samples from postnatal Bmal1 KO and WT mice over the course of 24 h under dark-dark conditions, with 4 samples taken every 6 h in each genotype. (a) PCA plot of the samples’ top 500 genes found rhythmic in WT. (b) PCA plot of the 12,067 genes with p > 0.5 in WT. (c-d) RNA-seq was performed on liver samples from adipocyte-specific Bmal1 KO and WT. (c) Bmal1 expression levels in WT and KO. (d) PCA plot among top 500 genes found rhythmic in WT, after filtering out low-expressed genes (mean expression <3). Rhythmicity testing by JTK_CYCLE. Abbreviations: PCA = principal component analysis; KO = knock-out; WT = wild-type.