Practical considerations for machine learning-enabled discoveries in spatial transcriptomics

Area of Science:

• Molecular biology
• Computational biology
• Genomics

Background:

• Multicellular development relies on precise spatial molecular patterns.
• Advanced imaging like spatial transcriptomics (ST) offers new insights into these patterns.
• Large ST datasets necessitate sophisticated computational tools for analysis.

Purpose of the Study:

• To highlight how machine learning (ML) can address key spatial transcriptomics (ST) analysis goals.
• To provide guidance on selecting appropriate ML tools for spatial biology data.
• To aid researchers in disentangling complex biological signals from noise.

Main Methods:

• Review of machine learning (ML) applications in spatial biology.
• Discussion of data science concepts relevant to ST data analysis.
• Presentation of heuristics for choosing ML tools.

Main Results:

• Identified specific ST analysis goals addressable by ML.
• Outlined four major data science concepts for tool selection.
• Provided practical heuristics for researchers.

Conclusions:

• ML is crucial for advancing spatial biology research using ST data.
• Understanding data science principles enhances the effective application of ML in ST.
• This work facilitates informed selection of computational tools for biological discovery.