AI + Proteomics Technology Services, Combining Dry and Wet Methods for Dual Assurance!
Release time:
2025-01-16
AI, as a revolutionary tool, has already demonstrated tremendous potential in the field of life sciences. In proteomics, machine learning and deep learning are widely used in protein-protein interactions, protein-nucleic acid interactions, and protein-small molecule docking.The release of AIphaFold3 in 2024 has significantly improved the accuracy of predicting protein-protein and protein-nucleic acid interactions compared to the previous version, AIphaFold2.

However, despite this, digital library screening can only serve as an effective tool to narrow down the scope for downstream validation. It often requires additional verification experiments from other systems to provide strong evidence for publications. Common methods include in vitro validation of interactions such as Pull-Down, EMSA, and yeast single/double hybrid one-on-one validation; as well as in vivo validation of interactions through immunoprecipitation experiments.
By combining digital library screening with validation experiments, sequences of interest can be selected for downstream verification, which not only improves the positivity rate but also reduces the workload of validation and enhances experimental efficiency.
Case Study

The research team led by Danesh Moazed at Harvard Medical School and the Howard Hughes Medical Institute (HHMI), in collaboration with the research group of Li Qing at Peking University, published a paper titled A replisome-associated histone H3-H4 tetramer chaperone required for epigenetic inheritance in Cell. The study utilized AlphaFold-Multimer to reveal that the DNA replisome component Mrc1/CLASPIN, which mediates the DNA replication checkpoint pathway, contains a region that is unrelated to replication stress signaling but is conserved in eukaryotes, which binds to the histone H3-H4 tetramer.
To validate AlphaFold's structural prediction, the histone-binding region of Mrc1 was confirmed in vitro using GST pull-down, where a single Mrc1 histone-binding region binds to an H3-H4 tetramer.

Figure: AlphaFold-Multimer predicts a conserved evolutionary region of Mrc1/CLASPIN that binds to the histone H3-H4 tetramer.

Figure: Pull-Down in vitro experiment validating the AlphaFold-Multimer prediction.
Reference 2

The article published in PLOS BIOLOGY in 2024 used AlphaFold computational predictions to reveal a physical interaction between Arl2 and Cdk5rap2. The study further validated this prediction using downstream techniques, including immunoprecipitation (Co-IP) and proximity ligation assays.

Figure: Digital prediction score ipTM and validation results. The interaction score ipTM of the known Arl2-interacting protein TBCD predicted through digital screening is 0.89, while the interaction score ipTM between Cdk5rap2 and Arl2 is 0.57.
Service Introduction
Digital Library Screening
Nanjing Pronetbio's digital library screening technology utilizes a self-designed multi-model refinement-based PPI prediction method. It integrates multiple novel protein AI tools and predicts protein interactions based on the 3D structure of proteins.


Digital Screening Service Flowchart for Protein-Protein and Protein-Nucleic Acid Interactions
The program undergoes three rounds of screening, with the following advantages:
1) High-throughput Screening: Self-developed program for batch screening, with a fast turnaround time.
2) Three Rounds of Screening: Multiple rounds of screening ensure accuracy.
3) Solving Pain Points: Addresses challenges such as proteins that are difficult or unable to be expressed.
4) Cost-effective: Reduces lab resources and costs, especially during the initial screening phase, avoiding unnecessary experiments.
5) Additional Services: Assistance in providing methodological descriptions for the methods section of publications.
Related News
2025-07-10
2025-07-08
2025-07-03
Literature Sharing | Regulation of co-translational mRNA decay by PAP and DXO1 in Arabidopsis
2025-07-01
2025-06-27
2025-06-24
2025-06-20