SNP密度对亲缘关系推断效能的影响及IBS算法的机器学习优化

doi:10.16288/j.yczz.25-287

遗传 ›› 2026, Vol. 48 ›› Issue (6): 570-588.doi: 10.16288/j.yczz.25-287

SNP密度对亲缘关系推断效能的影响及IBS算法的机器学习优化

戴律¹^,²(), 汤子琛³, 贾镇¹^,², 江丽², 赵传桐¹^,², 赵志远¹^,², 赵雯婷²(), 李彩霞¹^,²()

¹ 中国人民公安大学侦查学院，北京 100038
² 公安部鉴定中心，北京 100038
³ 江苏师范大学生命科学学院，徐州 221116

收稿日期:2025-12-08 修回日期:2026-02-05 出版日期:2026-06-20 发布日期:2026-02-12
通讯作者: 赵雯婷，博士，副主任法医师，研究方向：法医遗传学。E-mail: wtzhao@sibs.ac.cn;
李彩霞，博士，主任法医师，研究方向：法医遗传学。E-mail: licaixia@tsinghua.org.cn
作者简介:戴律，硕士研究生，专业方向：刑事科学技术。E-mail: 1130039162@qq.com
基金资助:
中央级公益性科研院所基本科研业务费专项资金(2024JB043);中央级公益性科研院所基本科研业务费专项资金(2024JB026);国家自然科学基金项目(82171870);北京市科技新星计划(20220484149)

SNP density impact on kinship inference and IBS-machine learning optimization

Lv Dai¹^,²(), Zichen Tang³, Zhen Jia¹^,², Li Jiang², Chuantong Zhao¹^,², Zhiyuan Zhao¹^,², Wenting Zhao²(), Caixia Li¹^,²()

¹ School of Criminal Investigation, People’s Public Security University of China, Beijing 100038, China
² Beijing Engineering Research Center of Crime Scene Evidence Examination, National Engineering Laboratory for Forensic Science, Institute of Forensic Science, Beijing 100038, China
³ School of Life Sciences, Jiangsu Normal University, Xuzhou 221116, China

Received:2025-12-08 Revised:2026-02-05 Published:2026-06-20 Online:2026-02-12
Supported by:
Fundamental Research Funds for Institute of Forensic Science(2024JB043);Fundamental Research Funds for Institute of Forensic Science(2024JB026);National Natural Science Foundation of China(82171870);Beijing Nova Program of Science and Technology(20220484149)

摘要/Abstract

摘要：

近年来，法医遗传领域报道了多个包含不同数量单核苷酸多态性(single nucleotide polymorphism，SNP)组合(panel)用于亲缘关系推断，但SNP位点数量对推断效能的影响及机器学习算法的应用缺乏系统探索。为此，本研究评估了SNP位点数量对亲缘关系推断效能的影响及机器学习方法对状态一致性(identity-by-state，IBS)算法的优化效果。首先，构建位点数量在15,476~20,838范围内的多个SNP panel，基于模拟家系评估似然比法和IBS算法在不同位点数量下的亲缘关系推断效能。在筛选出最优SNP panel后，利用真实家系进行验证，并进一步将IBS算法与机器学习方法结合以提升推断效能。结果显示，似然比法在六级和七级亲缘关系推断中的灵敏度与位点数量呈显著正相关。IBS算法四至七级亲缘关系推断的灵敏度虽然与位点数量呈显著正相关，但实际提升幅度有限(仅提升0.5%~2.2%)。基于上述结果，本研究确定了包含20,838个SNP位点的最优panel(21K panel)。21K panel基于似然比法可准确推断六级以内亲缘关系(六级亲缘关系推断灵敏度为93.65%)；基于IBS算法可准确推断三级以内亲缘关系(三级亲缘关系推断灵敏度为86.79%)。IBS算法结合机器学习后，四级亲缘关系推断灵敏度从69.10%提升至87.66%，五级和六级亲缘关系推断灵敏度分别从38.03%和21.41%提升至48.75%和37.80%。

关键词: 亲缘关系推断, 似然比法, IBS算法, 机器学习

Abstract:

In recent years, multiple panels containing varying numbers of single nucleotide polymorphisms (SNPs) have been reported in forensic genetics for kinship inference. However, systematic exploration of the impact of SNP number on inference performance and the application of machine learning algorithms remains lacking. Therefore, we evaluated the impact of SNP number on kinship inference performance and the optimization effects of machine learning methods on the identity-by-state (IBS) algorithm. We constructed multiple SNP panels with SNP numbers ranging from 15,476 to 20,838, and evaluated the performance of the likelihood ratio (LR) method and the IBS algorithm for kinship inference under different SNP numbers based on simulated pedigrees. After selecting the optimal SNP panel, we validated it using real pedigrees and further combined the IBS algorithm with machine learning methods to enhance inference performance. Our results showed that for the LR method, the sensitivity in inferring sixth and seventh degree kinships exhibited a significant positive correlation with SNP number. For the IBS algorithm, although the sensitivity in inferring fourth to seventh degree kinships showed a significant positive correlation with SNP number, the actual improvement was limited (only 0.5%~2.2% increase). Based on these results, we determined the optimal panel containing 20,838 SNPs (21K panel). The 21K panel based on the LR method could accurately infer kinships within sixth degree (with a sensitivity of 93.65% for sixth degree kinship inference), and the 21K panel based on the IBS algorithm could accurately infer kinships within third degree (with a sensitivity of 86.79% for third degree kinship inference). After combining the IBS algorithm with machine learning, the sensitivity for fourth degree kinship inference improved from 69.10% to 87.66%, the sensitivities for fifth and sixth degree kinships improved from 38.03% and 21.41% to 48.75% and 37.80%, respectively.

Key words: kinship inference, likelihood ratio method, IBS method, machine learning

戴律, 汤子琛, 贾镇, 江丽, 赵传桐, 赵志远, 赵雯婷, 李彩霞. SNP密度对亲缘关系推断效能的影响及IBS算法的机器学习优化[J]. 遗传, 2026, 48(6): 570-588.

Lv Dai, Zichen Tang, Zhen Jia, Li Jiang, Chuantong Zhao, Zhiyuan Zhao, Wenting Zhao, Caixia Li. SNP density impact on kinship inference and IBS-machine learning optimization[J]. Hereditas(Beijing), 2026, 48(6): 570-588.

图/表 15

表1

图1

图2

图3

表2

图4

图5

表3

表4

表5

表6

表7

图6

图7

表8

参考文献 31

[1]	Phillips C. The golden state killer investigation and the nascent field of forensic genealogy. Forensic Sci Int Genet, 2018, 36: 186-188. pmid: 30041097
[2]	Tillmar A, Kling D. Comparative study of statistical approaches and SNP panels to infer distant relationships in forensic genetics. Genes, 2025, 16(2): 114. pmid: 40004443
[3]	Jäger AC, Alvarez ML, Davis CP, Guzmán E, Han Y, Way L, Walichiewicz P, Silva D, Pham N, Caves G, Bruand J, Schlesinger F, Pond SJK, Varlaro J, Stephens KM, Holt CL. Developmental validation of the MiSeq FGx forensic genomics system for targeted next generation sequencing in forensic DNA casework and database laboratories. Forensic Sci Int Genet, 2017, 28: 52-70. pmid: 28171784
[4]	Gorden EM, Greytak EM, Sturk-Andreaggi K, Cady J, McMahon TP, Armentrout S, Marshall C. Extended kinship analysis of historical remains using SNP capture. Forensic Sci Int Genet, 2022, 57: 102636. pmid: 34896972
[5]	De Vries JH, Kling D, Vidaki A, Arp P, Kalamara V, Verbiest MMPJ, Piniewska-Róg D, Parsons TJ, Uitterlinden AG, Kayser M. Impact of SNP microarray analysis of compromised DNA on kinship classification success in the context of investigative genetic genealogy. Forensic Sci Int Genet, 2022, 56: 102625. pmid: 34753062
[6]	Tillmar A, Sturk-Andreaggi K, Daniels-Higginbotham J, Thomas JT, Marshall C. The FORCE panel: an all-in-one SNP marker set for confirming investigative genetic genealogy leads and for general forensic applications. Genes (Basel), 2021, 12(12): 1968. pmid: 34946917
[7]	Kling D, Tillmar A. Forensic genealogy—a comparison of methods to infer distant relationships based on dense SNP data. Forensic Sci Int Genet, 2019, 42: 113-124. pmid: 31302460
[8]	Kling D, Phillips C, Kennett D, Tillmar A. Investigative genetic genealogy: current methods, knowledge and practice. Forensic Sci Int Genet, 2021, 52: 102474. pmid: 33592389
[9]	Zhou Y, Browning SR, Browning BL. A fast and simple method for detecting identity-by-descent segments in large-scale data. Am J Hum Genet, 2020, 106(4): 426-437. pmid: 32169169
[10]	Seidman DN, Shenoy SA, Kim M, Babu R, Woods IG, Dyer TD, Lehman DM, Curran JE, Duggirala R, Blangero J, Williams AL. Rapid, phase-free detection of long identity- by-descent segments enables effective relationship classification. Am J Hum Genet, 2020, 106(4): 453-466. pmid: 32197076
[11]	Dou JZ, Sun BL, Sim XL, Hughes JD, Reilly DF, Tai ES, Liu JJ, Wang CL. Estimation of kinship coefficient in structured and admixed populations using sparse sequencing data. PLoS Genet, 2017, 13(9): e1007021. pmid: 28961250
[12]	Morling N, Allen RW, Carracedo A, Geada H, Guidet F, Hallenberg C, Martin W, Mayr WR, Olaisen B, Pascali VL, Schneider PM, Paternity Testing Commission of the International Society of Forensic Genetics. Paternity testing commission of the international society of forensic genetics: recommendations on genetic investigations in paternity cases. Forensic Sci Int, 2002, 129(3): 148-157. pmid: 12372685
[13]	Abecasis GR, Wigginton JE. Handling marker-marker linkage disequilibrium: pedigree analysis with clustered markers. Am J Hum Genet, 2005, 77(5): 754-767. pmid: 16252236
[14]	Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MAR, Bender D, Maller J, Sklar P, De Bakker PIW, Daly MJ, Sham PC. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet, 2007, 81(3): 559-575. pmid: 17701901
[15]	Zeng K, Zhao WT, Fang ZX, Li J, Liu J, Zhao D, Zhu BF, Li CX. Development and validation of a capture sequencing panel containing 9000 SNPs for inferring distant relatives in East Asian populations. Forensic Sci Int Genet, 2026, 81: 103341. pmid: 40845711
[16]	Abecasis GR, Cherny SS, Cookson WO, Cardon LR. Merlin—rapid analysis of dense genetic maps using sparse gene flow trees. Nat Genet, 2002, 30(1): 97-101. pmid: 11731797
[17]	Wu RG, Chen H, Li R, Zang Y, Shen XF, Hao B, Wang QW, Sun HY. Pairwise kinship testing with microhaplotypes: can advancements be made in kinship inference with these markers? Forensic Sci Int, 2021, 325: 110875. pmid: 34166816
[18]	Cui W, Chen M, Yang Y, Cai MM, Lan Q, Xie T, Zhu BF. Applications of 1993 single nucleotide polymorphism loci in forensic pairwise kinship identifications and inferences. Forensic Sci Int Genet, 2023, 65: 102889. pmid: 37247510
[19]	Manichaikul A, Mychaleckyj JC, Rich SS, Daly K, Sale M, Chen WM.Robust relationship inference in genome-wide association studies. Bioinformatics, 2010, 26(22): 2867-2873. pmid: 20926424
[20]	Liu J, Wei YL, Yang L, Jiang L, Zhao WT, Li CX. Testing of two SNP array-based genealogy algorithms using extended Han Chinese pedigrees and recommendations for improved performances in forensic practice. Electrophoresis, 2023, 44(17-18): 1435-1445. pmid: 37501329
[21]	Chawla NV, Bowyer KW, Hall LO, Kegelmeyer WP. SMOTE: synthetic minority over-sampling technique. J Artif Intell Res, 2002, 16(1): 321-357.
[22]	Van Den Heuvel E, Zhan ZZ. Myths about linear and monotonic associations: Pearson’s r, Spearman’s ρ, and Kendall’s τ. Am Stat, 2022, 76(1): 44-52.
[23]	Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc B, 1995, 57(1): 289-300.
[24]	Tomczak M, Tomczak E. The need to report effect size estimates revisited. An overview of some recommended measures of effect size. Trends Sport Sci, 2014, 1(21): 19-25.
[25]	Lakens D. Calculating and reporting effect sizes to facilitate cumulative science: a practical primer for t-tests and ANOVAs. Front Psychol, 2013, 4: 863. pmid: 24324449
[26]	Hill WG, Weir BS. Variation in actual relationship as a consequence of Mendelian sampling and linkage. Genet Res, 2011, 93(1): 47-64. pmid: 21226974
[27]	Huff CD, Witherspoon DJ, Simonson TS, Xing JC, Watkins WS, Zhang YH, Tuohy TM, Neklason DW, Burt RW, Guthery SL, Woodward SR, Jorde LB. Maximum- likelihood estimation of recent shared ancestry (ERSA). Genome Res, 2011, 21(5): 768-774. pmid: 21324875
[28]	Speed D, Balding DJ. Relatedness in the post-genomic era: is it still useful? Nat Rev Genet, 2015, 16(1): 33-44. pmid: 25404112
[29]	Wei YF, Zhu Q, Wang HY, Cao YY, Li X, Zhang XK, Wang YF, Zhang J. Pairwise kinship inference and pedigree reconstruction using 91 microhaplotypes. Forensic Sci Int Genet, 2024, 72: 103090. pmid: 38968912
[30]	Tabangin ME, Woo JG, Martin LJ. The effect of minor allele frequency on the likelihood of obtaining false positives. BMC Proc, 2009, 3(Suppl 7): S41. pmid: 20018033
[31]	Thompson EA. Identity by descent: variation in meiosis, across genomes, and in populations. Genetics, 2013, 194(2): 301-326. pmid: 23733848

编辑推荐

Metrics

www.chinagene.cn
备案号：京ICP备09063187号-4
总访问:,今日访问:,当前在线:

亲缘关系	亲缘关系系数	推断标准	零IBD共享统计量	推断标准
MZ	$\frac{1}{2}$	>$\frac{1}{{2}^{3/2}}$	0	<0.1
PO	$\frac{1}{4}$	($\frac{1}{{2}^{5/2}}$,$\frac{1}{{2}^{5/2}}$)	0	<0.1
FS	$\frac{1}{4}$	($\frac{1}{{2}^{5/2}}$, $\frac{1}{{2}^{5/2}}$)	$\frac{1}{4}$	(0.1, 0.365)
2nd	$\frac{1}{8}$	($\frac{\text{1}}{{\text{2}}^{\text{7}/\text{2}}}$, $\frac{\text{1}}{{\text{2}}^{\text{7}/\text{2}}}$)	$\frac{1}{2}$	(0.365, 1$-\frac{\text{1}}{{\text{2}}^{\text{3}/\text{2}}}$)
3rd	$\frac{1}{16}$	($\frac{\text{1}}{{\text{2}}^{\text{9}/\text{2}}}$, $\frac{\text{1}}{{\text{2}}^{\text{9}/\text{2}}}$)	$\frac{3}{4}$	(1$-\frac{\text{1}}{{\text{2}}^{\text{3}/\text{2}}}$, 1$-\frac{\text{1}}{{\text{2}}^{\text{3}/\text{2}}}$)
4th	$\frac{1}{32}$	($\frac{\text{1}}{{\text{2}}^{\text{11}/\text{2}}}$, $\frac{\text{1}}{{\text{2}}^{\text{11}/\text{2}}}$)	$\frac{7}{8}$	(1$-\frac{\text{1}}{{\text{2}}^{\text{5}/\text{2}}}$, 1$-\frac{\text{1}}{{\text{2}}^{\text{5}/\text{2}}}$)
5th	$\frac{1}{64}$	($\frac{\text{1}}{{\text{2}}^{\text{13}/\text{2}}}$, $\frac{\text{1}}{{\text{2}}^{\text{13}/\text{2}}}$)	$\frac{15}{16}$	(1$-\frac{\text{1}}{{\text{2}}^{\text{7}/\text{2}}}$, 1$-\frac{\text{1}}{{\text{2}}^{\text{7}/\text{2}}}$)
6th	$\frac{1}{128}$	($\frac{\text{1}}{{\text{2}}^{\text{15}/\text{2}}}$, $\frac{\text{1}}{{\text{2}}^{\text{15}/\text{2}}}$)	$\frac{31}{32}$	(1$-\frac{\text{1}}{{\text{2}}^{\text{9}/\text{2}}}$, 1$-\frac{\text{1}}{{\text{2}}^{\text{9}/\text{2}}}$)
7th	$\frac{1}{256}$	($\frac{\text{1}}{{\text{2}}^{\text{17}/\text{2}}}$, $\frac{\text{1}}{{\text{2}}^{\text{17}/\text{2}}}$)	$\frac{63}{64}$	(1$-\frac{\text{1}}{{\text{2}}^{\text{11}/\text{2}}}$, 1$-\frac{\text{1}}{{\text{2}}^{\text{11}/\text{2}}}$)
UN	0	≤0	1	1

SNP panel	亲缘关系	最小值	最大值	中位数	四分位差
15,476 SNPs	FS	1,072.95	3,024.46	1,947.54	345.67
	2nd	198.14	868.27	483.15	105.27
	3rd	41.03	432.76	214.26	77.44
	4th	2.14	221.04	80.55	38.53
	5th	-2.73	137.78	33.42	25.07
	6th	-1.89	91.12	13.44	15.10
	7th	-1.01	67.31	4.65	9.13
16,335 SNPs	FS	1,092.61	3,055.49	2,029.01	372.75
	2nd	202.07	796.77	505.16	112.90
	3rd	53.98	497.24	225.83	82.32
	4th	1.54	225.36	85.95	40.29
	5th	-2.53	152.54	36.09	25.72
	6th	-1.71	92.45	14.37	16.23
	7th	-0.96	52.27	5.10	9.34
17,190 SNPs	FS	1,124.99	3,391.81	2,123.87	382.52
	2nd	212.60	832.31	528.93	114.69
	3rd	43.67	498.70	235.10	84.76
	4th	2.72	257.00	90.38	42.63
	5th	-3.13	132.66	38.51	27.04
	6th	-1.65	104.99	15.60	17.00
	7th	-0.97	60.74	5.58	10.12
18,113 SNPs	FS	1,190.91	3,388.22	2,203.48	393.76
	2nd	208.09	862.78	551.54	120.46
	3rd	20.75	512.22	246.60	88.67
	4th	5.86	244.06	95.36	44.32
	5th	-2.74	149.03	41.03	29.33
	6th	-1.52	126.21	16.50	17.35
	7th	-0.86	54.70	6.02	10.36
19,022 SNPs	FS	1,251.28	3,499.48	2,302.01	407.16
	2nd	244.14	914.15	577.05	125.82
	3rd	55.03	537.41	259.37	91.19
	4th	9.56	258.09	101.22	46.68
	5th	-0.97	160.18	43.36	30.33
	6th	-1.57	109.03	17.75	18.81
	7th	-0.77	68.84	6.60	11.27
19,934 SNPs	FS	1,257.21	3,610.17	2,371.70	421.40
	2nd	235.84	962.51	600.05	127.67
	3rd	78.03	587.94	269.73	95.92
	4th	11.21	322.68	106.23	48.72
	5th	-2.28	172.12	45.85	31.39
	6th	-1.46	106.58	18.58	19.38
	7th	-0.66	72.88	6.90	11.54
20,838 SNPs	FS	1,259.09	3,726.38	2,466.37	445.04
	2nd	248.86	967.85	622.00	133.54
	3rd	63.64	620.12	279.62	97.89
	4th	1.51	297.68	111.32	51.26
	5th	-2.49	177.54	47.33	32.46
	6th	-1.26	116.53	19.80	20.38
	7th	-0.62	86.17	7.24	12.06

亲缘关系	t	灵敏度(%)	假阴性(%)	特异度(%)	假阳性(%)
FS/unrelated	4	100.00	0.00	100.00	0.00
2nd/unrelated	4	100.00	0.00	100.00	0.00
3rd/unrelated	4	100.00	0.00	99.90	0.10
4th/unrelated	4	100.00	0.00	99.80	0.20
5th/unrelated	4	99.56	0.44	99.50	0.50
6th/unrelated	4	93.65	6.35	99.30	0.70
	3	95.49	4.51	99.20	0.80
	2	96.72	3.28	98.30	1.70
	1	97.85	2.15	95.40	4.60
7th/unrelated	4	68.70	31.30	99.30	0.70
	3	74.24	25.76	98.90	1.10
	2	79.32	20.68	98.10	1.90
	1	86.21	13.79	93.60	6.40

真实亲缘关系	推断亲缘关系									灵敏度(%)	假阴性(%)	特异度(%)	假阳性(%)
真实亲缘关系	PO	FS	2nd	3rd	4th	5th	6th	7th	UN	灵敏度(%)	假阴性(%)	特异度(%)	假阳性(%)
PO	76	3	165							31.00	0.00
FS		131								100.00	0.00
2nd		7	316	9	1					94.89	0.00
3rd			3	300	136					68.34	0.00
4th				46	549	7				91.20	0.00
5th				3	413	469	28	1	1	51.54	0.11
6th					16	566	312	58	24	31.97	2.46
7th					5	126	407	226	121	35.70	13.67
UN						1	22	139	838			83.80	16.20

真实亲缘关系	推断亲缘关系										灵敏度(%)	假阴性(%)	特异度(%)	假阳性(%)
真实亲缘关系	PO	FS	2nd	3rd	4th	5th	6th	7th	>7th	UN	灵敏度(%)	假阴性(%)	特异度(%)	假阳性(%)
PO	244										100.00	0.00
FS		131									100.00	0.00
2nd			323	10							97.00	0.00
3rd			6	381	51	1					86.79	0.00
4th				36	416	123	13	7	2	5	69.10	0.83
5th					116	348	232	76	59	84	38.03	9.18
6th					13	148	209	142	159	305	21.41	31.25
7th					1	44	95	113	136	496	12.77	56.05
UN						17	64	97	133	689			68.90	31.10

SNP密度对亲缘关系推断效能的影响及IBS算法的机器学习优化

SNP density impact on kinship inference and IBS-machine learning optimization

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 15

参考文献 31

相关文章 11

编辑推荐

Metrics

IBS-ML	评估参数	PO (%)	FS (%)	2nd (%)	3rd (%)	4th (%)	5th (%)	6th (%)	7th (%)	UN (%)
IBS-XGBoost	灵敏度	100.0±0.0	100.0±0.0	97.0±2.7	83.2±7.3	81.2±5.2	47.6±11.7	35.7±13.0	14.0±9.0
	假阴性	0.0±0.0	0.0±0.0	0.0±0.0	0.0±0.0	0.0±0.0	5.5±2.7	33.9±11.2	44.9±9.9
	特异度									77.3±10.5
	假阳性									22.7±10.5
IBS-LightGBM	灵敏度	100.0±0.0	99.8±0.6	97.5±2.4	85.2±5.0	76.3±7.9	46.8±8.6	25.3±8.7	15.1±6.8
	假阴性	0.0±0.0	0.0±0.0	0.0±0.0	0.0±0.0	0.0±0.0	7.5±5.0	32.0±8.1	47.7±2.6
	特异度									74.6±16.2
	假阳性									25.4±16.2
IBS-RF	灵敏度	100.0±0.0	100.0±0.0	97.5±2.4	83.3±4.6	80.5±9.9	48.8±10.9	39.3±10.8	9.6±6.7
	假阴性	0.0±0.0	0.0±0.0	0.0±0.0	0.0±0.0	0.0±0.0	5.4±5.2	30.8±8.0	46.5±6.6
	特异度									76.3±11.7
	假阳性									23.7±11.7
IBS-DT	灵敏度	100.0±0.0	100.0±0.0	93.1±1.2	82.1±4.8	73.2±9.8	55.2±10.4	45.0±5.9	1.8±1.8
	假阴性	0.0±0.0	0.0±0.0	0.0±0.0	0.0±0.0	0.3±0.3	7.1±5.4	28.5±4.3	52.2±2.6
	特异度									72.3±13.5
	假阳性									27.7±13.5
IBS-KNN	灵敏度	100.0±0.0	100.0±0.0	97.2±2.6	85.2±7.3	74.3±8.1	44.1±5.5	25.0±6.3	24.5±3.4
	假阴性	0.0±0.0	0.0±0.0	0.0±0.0	0.0±0.0	0.3±0.3	9.0±2.3	20.7±6.8	33.3±13.0
	特异度									46.5±14.8
	假阳性									53.5±14.8
IBS-SVM	灵敏度	100.0±0.0	99.4±1.5	97.3±2.5	87.6±4.8	61.7±7.8	49.4±11.0	44.3±10.4	40.1±22.4
	假阴性	0.0±0.0	0.0±0.0	0.0±0.0	0.0±0.0	0.0±0.0	2.2±2.0	8.7±7.2	15.0±12.9
	特异度									26.5±22.3
	假阳性									73.5±22.3

[1]	李志宏, 陈淼琪, 袁晓珺, 黄华君, 黄琬婷, 周飘雁, 曾晨, 冯许诺, 杨洛瑶, 黄树强, 谭翠钰, 陈彩蓉, 颜秋霞. 基于microRNA与生物信息学筛选非梗阻性无精子症的生物标志物[J]. 遗传, 2026, 48(3): 301-312.
[2]	梁卉, 王雪, 司敬方, 张毅. 利用基因组标记和机器学习算法对中国牛品种的分类准确性研究[J]. 遗传, 2024, 46(7): 530-539.
[3]	章子怡, 王棨临, 张俊有, 段迎迎, 刘家欣, 刘赵硕, 李春燕. 多组学数据驱动的机器学习模型在乳腺癌生存及治疗响应预测中的应用[J]. 遗传, 2024, 46(10): 820-832.
[4]	郑慧怡, 吴华煊, 杜志强. 肠道宏基因组图像增强和深度学习改善代谢性疾病分类预测精度[J]. 遗传, 2024, 46(10): 886-896.
[5]	陈栋, 王书杰, 赵真坚, 姬祥, 申琦, 余杨, 崔晟頔, 王俊戈, 陈子旸, 王金勇, 郭宗义, 吴平先, 唐国庆. 基于机器学习的猪生长性状基因组预测[J]. 遗传, 2023, 45(10): 922-932.
[6]	孔永强, 刘金凯, 顾佳琪, 徐景怡, 郑雨诺, 魏以梁, 伍少远. 南-北方汉族人、韩国人和日本人遗传划分机器学习模型优化方案[J]. 遗传, 2022, 44(11): 1028-1043.
[7]	胡雅丽, 戴睿, 刘永鑫, 张婧赢, 胡斌, 储成才, 袁怀波, 白洋. 水稻典型品种日本晴和IR24根系微生物组的解析[J]. 遗传, 2020, 42(5): 506-518.
[8]	赵学彤, 杨亚东, 渠鸿竹, 方向东. 组学时代下机器学习方法在临床决策支持中的应用[J]. 遗传, 2018, 40(9): 693-703.
[9]	张桂珊, 杨勇, 张灵敏, 戴宪华. 机器学习方法在CRISPR/Cas9系统中的应用[J]. 遗传, 2018, 40(9): 704-723.
[10]	彭哲也,唐紫珺,谢民主. 机器学习方法在基因交互作用探测中的研究进展[J]. 遗传, 2018, 40(3): 218-226.
[11]	侯妍妍，应晓敏，李伍举. microRNA计算发现方法的研究进展[J]. 遗传, 2008, 30(6): 687-696.