notesum.ai

Published at November 13

MVKTrans: Multi-View Knowledge Transfer for Robust Multiomics Classification

cs.LG

cs.AI

Released Date: November 13, 2024

Authors: Shan Cong¹, Zhiling Sang², Hongwei Liu³, Haoran Luo¹, Xin Wang⁴, Hong Liang³, Jie Hao⁵, Xiaohui Yao¹

Aff.: ¹Harbin Engineering University, Harbin, China; Qingdao Innovation and Development Center, Harbin Engineering University, Qingdao, China; ²Qingdao Innovation and Development Center, Harbin Engineering University, Qingdao, China; ³Harbin Engineering University, Harbin, China; ⁴Department of Surgery, Chinese University of Hong Kong, Hong Kong, China; ⁵Institute of Medical Information, Chinese Academy of Medical Sciences & Peking Union Medical College, Beijing, China

Arxiv: http://arxiv.org/abs/2411.08703v1

Method	Fusion	ROSMAP			LGG
Method		ACC (%)	F1 (%)	AUC (%)	ACC (%)	F1 (%)	AUC (%)
SVM	early	77.0±2.4	77.8±1.6	77.0±2.6	75.4±4.6	75.7±5.0	75.4±4.6
XGBoost		76.0±4.6	77.2±4.5	83.7±3.0	75.6±4.0	76.7±3.2	84.0±2.3
NN		75.5±2.1	76.4±2.1	82.7±2.5	73.7±2.3	74.8±2.4	81.0±3.7
GRridge	intermediate	76.0±3.4	76.9±2.9	84.1±2.3	74.6±3.8	75.6±3.6	82.6±4.4
BSPLSDA		75.3±3.3	76.4±3.5	83.8±2.1	68.5±2.7	66.2±3.0	73.0±2.6
CF		78.4±1.1	78.8±0.5	88.0±0.5	81.1±1.2	82.2±0.4	88.1±0.4
GMU		77.6±2.5	78.4±1.6	86.9±1.6	80.3±1.5	80.8±1.2	88.6±1.2
Mogonet	correlation	81.5±2.3	82.1±2.2	87.4±1.2	81.6±1.6	81.4±1.4	84.0±2.7
MODLIM	correlation	84.3±1.2	85.0±0.8	89.1±1.2	82.8±0.7	82.5±2.3	86.1±1.5
Dynamics	trustworthy	84.2±1.3	84.6±0.7	91.2±0.7	83.3±1.0	83.7±0.4	88.5±0.4
MVKTrans (ours)	distillation	86.4±0.5^∗	86.4±0.4^∗	92.5±0.9^∗	84.7±0.3^∗	84.7±0.7^∗	91.4±1.1^∗
Method	Fusion	BRCA			KIPAN
Method		ACC (%)	F1-w (%)	F1-m (%)	ACC (%)	F1-w (%)	F1-m (%)
SVM	early	72.9±1.8	70.2±1.5	64.0±1.7	99.5±0.3	99.5±0.3	99.4±0.4
XGBoost		78.1±0.8	76.4±1.0	70.1±1.7	99.3±0.8	99.3±0.8	98.9±1.4
NN		75.4±2.8	74.0±3.4	66.8±4.7	99.1±0.5	99.1±0.5	99.1±0.5
GRridge	intermediate	74.5±1.6	72.6±1.9	65.6±2.5	99.4±0.4	99.4±0.4	99.3±0.4
BSPLSDA		63.9±0.8	52.2±1.6	35.1±2.2	91.9±1.2	91.8±1.3	89.5±1.4
CF		81.5±0.8	81.5±0.9	77.1±0.9	99.2±0.5	99.2±0.5	99.2±0.5
GMU		80.0±3.9	79.8±5.8	74.6±5.8	99.2±0.5	99.2±0.5	98.8±0.9
Mogonet	correlation	82.9±1.8	82.5±1.6	77.4±1.7	99.9±0.2	99.9±0.2	99.9±0.2
MODLIM	correlation	84.5±0.9	84.0±1.6	80.4±1.2	99.2±0.8	99.2±0.8	99.2±0.8
Dynamics	trustworthy	87.7±0.3	88.0±0.5	84.5±0.5	99.9±0.2	99.9±0.2	99.9±0.2
MVKTrans (ours)	distillation	88.6±0.2^∗	88.9±0.1^∗	86.3±0.2^∗	99.9±0.2	99.9±0.2	99.9±0.2