ドロップダウンアイコン
ドロップダウンアイコン
ドロップダウンアイコン
JP
/
EN
メニューアイコン
Research
Craif Reports a Machine Learning Method for Correcting Run to Run Variation in Body Fluid microRNA Sequencing, Presented at the 9th Liquid Biopsy Study Group
2025.03.17

A methods presentation on reducing measurement variation in body fluid microRNA analysis. The work covers body fluids generally rather than urine alone, and no clinical performance figures were reported.

TOKYO, March 16, 2025 - Craif Inc., a bio-AI company developing a urine analysis platform for early detection of cancer and other diseases, said it presented a methods study at the 9th Liquid Biopsy Study Group, held February 7 and 8, 2025, in Shinjuku, Tokyo. The presentation, titled "Batch Effect Correction in Body Fluid miRNA Analysis Using Small RNA-seq," describes a data processing method intended to reduce the run-to-run measurement variation, known as batch effect, that arises when microRNA in body fluids is measured by small RNA sequencing.

This was a methods presentation. No sensitivity, specificity, AUC, cohort or patient result was disclosed, and no clinical claim is made. The work has not been published in a peer-reviewed journal.

What Craif says the method does

  • Reduces measurement variation between analytical runs. Craif developed a new correction method for run-to-run error, or batch effect, in small RNA sequencing of microRNA in body fluids. Craif expects this to make high-accuracy analysis of body fluid microRNA more stable.
  • Extends to data not seen before. By combining repeated measurements of the same specimen with machine learning, the correction can be applied to newly acquired data rather than only to datasets already in hand. Craif expects this to hold accuracy steadier across the range of specimens seen in clinical practice.
  • Addresses a problem the whole field has. Quality control of measurement batches in nucleic acid analysis is difficult and matters across the industry. Craif expects its method to advance that problem and support a platform for machine learning-based liquid biopsy analysis.

Why batch effects matter

MicroRNA in body fluids such as blood and urine has drawn attention as a route to early cancer detection that does not hurt the patient. Because body fluid microRNA is present in very small amounts, small errors introduced by each measurement run cannot be ignored. Existing methods can correct batch effects in already collected data, but applying appropriate correction to newly acquired data is sometimes difficult, making accuracy and test yield hard to keep stable.

In this study, Craif combined repeated measurements of the same specimen with machine learning to build a batch effect correction that also handles previously unseen data. Craif expects the resulting reduction in measurement error to support further gains in the accuracy of body fluid biomarker analysis for early cancer detection.

The miSignal series is not a medical device. It determines risk by statistical calculation from analyzed information. It does not substitute for a medical diagnosis of whether a person has cancer, and a low risk result does not mean cancer is absent or that cancer will not occur in the future.

About the meeting

About Craif

Craif is a bio-AI company founded in 2018 that develops non-invasive tests for the early detection of cancer. Its platform combines proprietary biomarker detection from urine and other bodily fluids with AI to assess disease risk. Building on one of the world's largest urine datasets, Craif aims to create a urine analysis platform that detects a wide range of diseases early, with cancer as the first step. The company is based in Tokyo and operates a wholly owned laboratory in San Diego, opened in 2026. Learn more at craif.com.

Media contact

press_us@craif.com