A Machine Learning Framework Identifies Plastid-Encoded Proteins Harboring C3 and C4 Distinguishing Sequence...

Area of Science:

• Plant biology
• Evolutionary biology
• Genomics

Background:

• C4 photosynthesis has evolved independently over 60 times, with a significant portion originating in the PACMAD clade of grasses.
• This evolutionary convergence makes the PACMAD clade a prime subject for studying the genomic underpinnings of C4 photosynthesis.

Purpose of the Study:

• To utilize machine learning (ML) to screen plastid genomes for C3/C4 distinguishing information within PACMAD species.
• To identify specific plastid genes and sequence sites that are highly predictive of photosynthetic type.

Main Methods:

• Employing a machine learning approach to analyze sequenced plastid genomes from PACMAD species.
• Training ML models on plastid-encoded protein sequences to classify C3 versus C4 photosynthesis.
• Identifying key genes and residues most predictive of C3/C4 status.

Main Results:

• Certain plastid-encoded protein sequences contain informative data for accurate ML-based C3/C4 classification.
• An RbcL-trained ML model achieved over 99% accuracy in distinguishing C3 and C4 photosynthetic types.
• Key predictive sequences include RbcL, NAD(P)H dehydrogenase subunits, and specific residues.

Conclusions:

• Plastid gene sequences hold significant information for predicting photosynthetic pathways, suggesting functional roles in C3/C4 metabolism.
• The ML framework successfully pinpointed critical genetic elements involved in C4 photosynthesis evolution and maintenance.
• This methodology can be broadly applied to investigate genotype-phenotype relationships in other biological systems.