Machine Learning Revolutionizes Microscopy

Advances in imaging, genomics, and other technologies have inundated the life sciences with data. When analyzing cells from the brain tissue of Alzheimer’s patients, biologists may need to explore various characteristics such as cell type, gene expression, location in the tissue, and more. While experimental probing can now occur with multiple measurements, data analysis often remains limited to one type of measurement at a time.

Dealing with “multimodal” data necessitates innovative computational tools, a domain where Xinyi Zhang excels.

As a fourth-year MIT PhD student, Zhang merges machine learning and biology to decipher fundamental biological principles, especially in areas where conventional methods fall short. Collaborating with researchers at the Broad Institute and beyond, Zhang, under MIT Professor Caroline Uhler’s mentorship, has initiated several endeavors aimed at constructing computational frameworks to unravel cell regulatory mechanisms.

Zhang envisions these efforts as stepping stones towards understanding how cells, tissues, and organs function, why diseases arise, and why some conditions are curable, whereas others are not.

During her leisure time, Zhang engages in equally ambitious activities, including sailing, skiing, ice skating, rock climbing, performing with MIT’s Concert Choir, and even flying single-engine planes. (She earned her pilot’s license in November 2022.)

“I enjoy exploring new places and endeavors,” she states humbly.

Uhler, her mentor, marvels at Zhang’s unassuming nature, revealing surprises in each interaction.

“Every time, you discover something like, ‘Oh, now she’s mastering flying,'” Uhler comments appreciatively. “Everything she undertakes stems from genuine passion. She strives for excellence in her pursuits, which is truly inspiring.”

Zhang’s fascination with biology sparked during her high school days in Hangzhou, China. Her curiosity was piqued by the challenges her biology teachers couldn’t immediately address, making it the most captivating subject for her.

Her initial interest in biology evolved into a passion for bioengineering. Encouraged by her parents, who were middle school educators, Zhang pursued this field in conjunction with electrical engineering and computer science at the University of California, Berkeley.

Graduating in 2020, Zhang was eager to commence MIT’s EECS PhD program, delayed due to the Covid-19 pandemic. Nonetheless, in December 2022, she, alongside Uhler and two collaborators, published a paper in Nature Communications.

The groundwork for the publication was laid by co-author Xiao Wang, who had previously collaborated with the Broad Institute to develop a unique spatial cell analysis technique. This method integrated various cell imaging and gene expression types for the same cell while mapping its location in the tissue sample, an unprecedented approach.

This innovation had broad applications, revolutionizing disease progression tracking, albeit challenging to analyze due to the extensive multimodal data generated. Enter Zhang, intrigued by creating a computational technique capable of handling this complexity.

Emphasizing chromatin staining as their imaging choice, known for cost-effectiveness and data richness, the team began integrating Wang’s spatial analysis methodologies. Zhang embarked on designing an autoencoder to facilitate this integration.

Autoencoders, neural networks that encode and compress extensive high-dimensional data, were employed to expand data dimensions in this scenario. Zhang’s autoencoder operated inversely, elevating input data dimensions. This facilitated cross-animal data combining and eradicated technical discrepancies irrelevant to biological variations.

The resulting technology, abbreviated as STACI, was used in the paper to elucidate Alzheimer’s disease progression observations with diverse spatial and imaging methods. Zhang believes this adaptable model can analyze a plethora of diseases effectively.

Dreaming of constructing a comprehensive human life model given ideal resources, Zhang acknowledges the limitations of time and resources. However, her ambition remains undaunted, desiring to tackle the most complex unanswered questions with her skills.

Currently finalizing neurodegeneration research involving frontal cortex imaging and predicting protein images from protein sequences and chromatin imaging, Zhang maintains her focus on addressing biologically significant questions to broaden understanding in uncharted territories.