scHiCluster is a single-cell clustering algorithm for Hi-C contact matrices. It is based on imputations using linear convolution and random walk. Please go to here to download it.


DCell is a neural network model of budding yeast, a basic eukaryotic cell. The model structure corresponds exactly to a hierarchy of 2,526 cellular subsystems. Given this neural network structure, DCell has been trained to translate genotype to phenotype. Given a genotype consisting of a set of deleted genes in S. cerevisiae, DCell will predict the cell growth phenotype and explain how this phenotype arises from changes in the states of cellular subsystems. Please go to here to download it.


The Data-Driven Ontology Toolkit (DDOT) facilitates the inference, analysis, and visualization of biological hierarchies using a data structure called an ontology. To download the package pleaese go to here. The resulted data-driven ontology can be viewed with our interactive viewer HiView.


NBS-ESP addresses the contamination problem caused by frequently mutated genes in network-based patient stratification in cancer genomics. To alleviate the noise, NBS-ESP only considers using selective gene interactions under evolutionary pressure rather than all human gene interactions in network propagation. The code can be downloaded at NBS-ESP.


NBS2 learns the mutated subnetworks underlying tumor subtypes using a supervised approach. Given an annotated molecular network and reference tumor mutation profiles for which subtypes have been predefined, NBS2 is trained by adjusting the weights on interaction features such that network propagation best recovers the provided subtypes. The code can be downloaded at NBS^2.


CNFpred is a new protein structure prediction program developed by our group, excelling at the alignment of distantly-related proteins with sparse sequence profile and that of a single target to multiple templates. In particular, CNFpred can generate much better alignments than our previous single-template threading program RaptorX. To predict structures for your protein sequences online, please click here. To download a standalone package please click here.


DeepAlign is a program for pairwise protein structure alignment. Different from many other structure alignment tools, DeepAlign aligns two protein structures using evolutionary information in addition to geometric similarity. Therefore, DeepAlign can align many more evolutionarily-related residues together than other tools. That is, the DeepAlign alignment makes more biological sense. Please download the binary code at DeepAlign_V1.00. You can also find a online version at here.


PDB2CLE is a program that can turn a PDB file to a sequence of structural alphabets. In paricular, each structural alphabet is a discretized set of representatives of the local protein structure. These structural alphabets are defined by clustering local fragments in protein structures through a variety of geometric meatures, as well as different segment length. Please download the binary code at here.


This is a software package that implemented our RECOMB 2014 work. It can do remote homology search by representing protein families as Markov Random Fields (MRF) and align two protein families by aligning two MRFs. By using MRFs, the long distance correlation between amino acids can be incorporated into remote homology detection. An ADMM-based method is applied so that the optimization can be done efficiently. Please download the binary code at here.


This is a web server that implemented our RECOMB 2015 work. Starting from a single protein sequence, It can predict protein contact map for the protein that integrates joint multi-family EC analysis and supervised learning, both of which are used to handle the case when not many sequence homologs are available. Different from existing single-family EC analysis that uses residue co-evolution information in only the target protein family, our joint EC analysis uses residue co-evolution in both the target family and its related families, which may have divergent sequences but similar folds. The web server address is here.


AcconPred is a software package that helps predicting solvent accessibility and contact number of a protein simultaneously. The method is based on a shared weight multi-task learning framework under a novel Conditional Neural Field (CNF) model. The multi-task learning framework on a collection of related tasks provides more accurate prediction than the framework trained only on a single task. The software can be found at here. The training and testing dataset can be found at here. The feature file can be at here. The NR database used by the prediction can be found at here.

Notice! For using our server RaptorX, registration is NOT required, but we STRONGLY RECOMMEND you to provide an email address in submission, which can be used to retrieve the results of all your jobs.