State-of-the-art in RNA 3D structure prediction
Link to article: 59940284 (biorxiv.org)
https://youtu.be/oe-w1Xx1p1g?si=5vGgS2PIwvqIgEaN
The Critical Assessment of Protein Structure Prediction (CASP) is a renowned event in the field of structural biology, where researchers from around the world compete to predict the structures of proteins. In recent years, CASP has expanded its scope to include RNA structure prediction, recognizing the growing importance of RNA in biological processes. In this blog post, we will summarize the main points from a YouTube video transcript where experts discussed the challenges and opportunities in assessing RNA structure predictions during CASP.
The Historic Opportunity:
CASPRNA, as it was referred to in the video transcript, offered a historic opportunity for the RNA field. The assessment team, composed of four experts, was excited about the potential to advance RNA structure prediction methods. Specifically, they aimed to encourage the development of deep learning-based approaches for RNA, as well as promote the use of cryo-electron microscopy (cryo-EM) for RNA structure determination.
Deep Learning and RNA:
One of the key hopes for CASPRNA was to push forward the application of deep learning methods in RNA structure prediction. The team presented examples where deep learning scores helped identify more accurate RNA structures among submitted models. However, there was room for improvement, as some structures still fell short of experimental data.
Cryo-EM and RNA:
Another major goal was to establish RNA structure assessment based on cryo-EM, a technique that had not been extensively applied to RNA until recently. The team showcased how cryo-EM had enabled the determination of RNA structures, setting records for solving small RNA molecules. CASP was seen as an ideal platform for assessing such targets due to its prior experience with similar low-resolution data.
Quantitative Assessment:
To gain insights into the quality of RNA structure predictions, the CASPRNA team developed quantitative assessment metrics. They introduced the GDT (Global Distance Test) and LDDT (Local Distance Difference Test) scores for RNA, inspired by their use in the protein structure assessment. These metrics aimed to evaluate both global and local structural accuracy.
Modeling RNA Flexibility:
RNA is known for its flexibility, and for some targets, multiple structural states are possible. CASPRNA addressed this challenge by allowing predictors to submit models representing different RNA states. The team assessed models against all available experimental structures, selecting the best-scoring ones for evaluation.
Results and Rankings:
The overall rankings in CASPRNA revealed that the top four predictors consistently outperformed others. Surprisingly, these groups had extensive experience in RNA research, rather than being newcomers driven by deep learning techniques. This raised questions about the influence of deep learning in RNA structure prediction, as it appeared that experienced researchers were leading the field.
Challenges with RNA-Protein Complexes:
While progress was evident in RNA-only targets, RNA-protein complexes posed significant challenges. Many predictions struggled to accurately capture the RNA-protein interactions and overall fold, highlighting areas for improvement.
Secondary Structure Assignment:
Assigning correct secondary structures was critical for accurate RNA structure prediction. In some cases, the availability of literature, chemical probing data, or RNA families aided in secondary structure determination. However, the intuitive understanding of RNA secondary structure often played a crucial role in improving predictions.
CASPRNA provided a significant boost to the assessment of RNA structure predictions, with valuable insights gained from quantitative metrics like GDT and LDDT. The competition highlighted the strengths and limitations of current methods, paving the way for future advancements in RNA structure prediction. While challenges remain, CASPRNA demonstrated the collaborative spirit and dedication of researchers in advancing our understanding of RNA structures.
- Parts of this article were written using Generative AI
- Subscribe/leave a comment if you want to stay up-to-date with the latest AI trends.
Plug: Checkout all my digital products on Gumroad here. Please purchase ONLY if you have the means to do so. Use code: MEDSUB to get a 10% discount!
