Skip to main content

MP3

MP3 is a transformer-based AI model for molecule programming, enabling the redesign, diversification, and de novo design of protein sequences. This advanced tool aids in protein engineering for applications such as drug development, enzyme creation, and therapeutic design. The terms protein engineering or optimization typically refer to small modifications to natural proteins, while protein design involves more extensive changes. De novo protein design implies creating a sequence from scratch, though it may still build on known templates. With its latest version, MP4, MP3 enhances the efficiency of protein design by leveraging machine learning techniques to manipulate amino acid sequences, making it essential for biotechnological advancements and personalized medicine.

Redesign A0A1L9RXX7 residues 40-60

design1

Inputs for Protein Sequence Design

  • Amino Acid Sequence: A single-chain protein sequence with a length of fewer than 300 amino acids.
  • Target Function (optional): A keyword describing the desired protein function, such as “leucine rich repeat” or “hydrolase.”
  • Temperature (optional): A parameter that influences the design; lower values introduce fewer changes, while higher values introduce more changes.

Outputs from MP3

  • Designed Protein Sequence: The optimized amino acid sequence generated based on the input parameters.
  • Confidence Scores: Metrics indicating the reliability of the predictions for the designed sequences.

MP3 Protein Design Example​s

Diversify Q7L266 with temperature 2.0

design3

Create a de novo hydrolase

design4

Create an asparagine protein

design5

Analyzing MP3 Results:

Analyzing MP3 results involves evaluating designed protein sequences and predicted functions. Key aspects include reviewing optimized amino acid sequences, assessing confidence scores, and comparing with template structures. This analysis informs potential applications and guides experimental validation.

Designed Protein Sequences

Review the optimized amino acid sequences generated by MP3 to assess their relevance to the intended functions. Consider how the modifications align with the desired characteristics and whether they reflect the objectives of the protein design process.

Predicted Protein Functions

Analyze the predicted functions associated with the designed sequences to determine how well they align with the desired biological roles. This involves evaluating whether the predictions support the intended applications, such as therapeutic uses or enzymatic activity.

Confidence Scores in Protein Design

Assess the confidence metrics provided for each prediction. Higher scores indicate more reliable designs, which can guide researchers in prioritizing certain sequences for experimental validation or further investigation.

Structural Models of Designed Proteins

Examine the generated structural models or visualizations to gain insights into how the designed sequences are likely to fold and interact with other molecules. Understanding these structural aspects can inform the potential functionality of the proteins.