Introduction
The formation of biomolecular condensates by liquid‒liquid phase separation has become a universal mechanism for spatiotemporal coordination of biological activities in cells and has been widely observed to directly regulate the key cellular processes involved in cancer cell pathology. However, the complexity of protein sequences and the diversity of conformations are inherently disordered, which poses great challenges for LLPS protein calculations and experimental research.
Herein, we proposed a novel predictor (PredLLPS_PSSM) for LLPS protein identification based only on sequence evolution information. PredLLPS_PSSM provides the option of predicting not only LLPS proteins but also PS-Self and PS-Part proteins. the source code for PredLLPS_PSSM are freely available at here.
Service
By submitting the data and your contact information, we will return the results to you by email.
- 01 Your email is the only way to communicate with you after we have completed the task you have submitted, so please be sure to fill it out carefully.
- 02 Your sequences must be submitted in fasta format and the CSV file of the AB PSSM features you need is from POSSUM. Please ensure that the fasta file submitted to POSSUM is the same as the fasta file submitted this time. And the example files are available at here.
- 03 The returned results will be returned with the descriptor as the identifier, so please use a unique descriptor for each sequence if possible.
- 04 The data you submit should meet the following conditions. 1. The submitted sequence length should be no less than 50 and no longer than 5000. 2. The number of sequences submitted is within 500. 3. Submit A sequence of one and only 20 kinds of amino acids, including 'A', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'K', 'L', 'M', 'N', 'P', 'Q', 'R', 'S', 'T', 'V', 'W', 'Y'.
Dataset
We built two new independent LLPS protein test datasets, which were collected from LLPSDB V2.0. We first filtered out proteins with posttranslational modification sites, invalid amino acids and lengths less than 50 or greater than 5000. Then, we weeded out those sequences with a similarity greater than 0.4 from the database of LLPS+, LLPS-, SaPS, and PdPS applying CD-HIT. Finally, there were 48 PS-Self proteins named Ind_Test_I and 140 PS-Part proteins named Ind_Test_II.
We collected LLPS proteins from the latest version of PhaSepDB V2.1, LLPSDB V2.0, and PhaSePro to construct the training data for the final prediction model. Through The CD-HIT algorithm with a minimum similarity threshold of 0.4 was used to remove proteins with high similarity. we obtained 578 LLPS proteins, 246 PS-self proteins, and 410 PS-part proteins. The same number of non-LLPS proteins (negative dataset) were randomly selected from PDB* correspondingly. For the convenience of description, we named the three final training datasets LLPS_Training, PS_Self_Training and PS_Part_Training, which are used to train the models.
In addition to the data in the ready-made database, we obtain nine protein sequences from Insect data that have been experimentally verified to undergo LLPS.
The dataset is available for download at the following
Contact Us
If you have any questions, please feel free to contact us.
Name:
Shengming Zhou
Email:
shengming_zhou7@163.com
Location:
No. 1, Linghai Road, Dalian City, Liaoning Province, China