Phage-host interactions are appealing systems to study co-evolution. Their roles in human health and diseases as well as novel therapeutics development also have been increasingly emphasized. Meanwhile, such interactions leave signals in bacterial and phage genomic sequences, defined as phage-host interaction signals (PHIS), allowing us to predict novel phage-host interactions. Due to the intrinsic complexity and recent emerging of metagenomics sequencing data, there is an urgent requirement to develop computational tools to analyze massive data and extract meaningful information. Here, we seize comprehensive in silico PHIS and utilize sophisticated bioinformatics to develop PHISDetector, a web tool to detect and systematically study diverse in silico PHIS, including analyses for co-occurrence/co-abundance patterns, oligonucleotide profile/sequence composition, CRISPR-targeting, prophages, phage genome similarity, protein-protein interactions, and special gene check. PHISDetector accepts various genomic and metagenomic data as input and provides well-designed visualizations and detailed data tables to download. Prediction tasks are processed remotely by the server using custom python scripts and a series of public tools. PHISDetector can be accessed at http://www.microbiome-bigdata.com/PHISDetector/index/.