# Pattern Recognition of Pyrolysis Bio-Oils by GC×GC-TOFMS with Tile-Based Feature Selection and Principal Component Analysis

## TL;DR

This study uses advanced analytical techniques to differentiate bio-oils from sugar cane bagasse and straw, revealing distinct chemical profiles that could guide fuel production methods.

## Contribution

The study introduces tile-based feature selection with GC×GC-TOFMS and PCA to identify subtle differences in bio-oil compositions.

## Key findings

- Bagasse bio-oil is rich in phenolics and hexose derivatives like furans and aldehydes.
- Straw bio-oil contains more hydrocarbons and fatty acid methyl esters.
- Tile-based FRA identified 16 differential features and low-intensity compounds not detected by traditional methods.

## Abstract

Chemometrics associated with advanced analytical separation
methods
are crucial for the chemical profiling of complex samples, such as
bio-oil, enabling more accurate and efficient identification of differential
features. The composition of bio-oils influences the selection of
pretreatment methods for fuel production, which may include processes
such as filtration, guard bed usage, or reactions such as hydrothermal
liquefaction and esterification. This study focuses on the chemical
profiling of pyrolytic bio-oils from sugar cane bagasse and straw
using comprehensive two-dimensional gas chromatography coupled with
time-of-flight mass spectrometry (GC×GC-TOFMS). Chemometric approaches
such as tile-based Fisher ratio analysis (FRA) and principal component
analysis (PCA) are employed for the feature selection of class-differentiating
analytes. Bio-oils from both feedstocks exhibited chromatographic
profiles with subtle differences, which were observed in the composition
and relative abundance of specific compound classes. Bagasse bio-oil
was rich in phenolics and hexose derivatives, such as furans and aldehydes.
In contrast, straw bio-oil presented a higher abundance of hydrocarbons
and fatty acid methyl esters. Tile-based FRA enabled the identification
of 16 differential features and the detection of low-intensity compounds,
such as long-chain esters and hydrocarbons, not previously detected
by the peak table-based approach. PCA based on these differential
features explained 98.7% of the total variance (PC1 + PC2), clearly
grouping bio-oils by feedstock origin. The findings highlight the
potential of GC×GC-TOFMS and chemometrics for differentiating
bio-oils, demonstrating the importance of advanced analytical techniques
in studying biomass conversion processes and characterizing bioproducts.