In cancer genomics, the mutually exclusive patterns of somatic mutations are important biomarkers that are suggested to be valuable in cancer diagnosis and treatment. However, detecting these patterns of mutation data is an NP-hard problem, which pose a great challenge for computational approaches. Existing approaches either limit themselves to pair-wise mutually exclusive patterns or largely rely on prior knowledge and complicated computational processes. Furthermore, the existing algorithms are often designed for genotype datasets, which may lose the information about tumor clonality, which is emphasized in tumor progression. In this paper, an algorithm for multiple sets with mutually exclusive patterns based on a fuzzy strategy to deal with real-type datasets is proposed. Different from the existing approaches, the algorithm focuses on both similarity within subsets and mutual exclusion among subsets, taking the mutual exclusion degree as the optimization objective rather than a constraint condition. Fuzzy clustering of the is done mutations by method of membership degree, and a fuzzy strategy is used to iterate the clustering centers and membership degrees. Finally, the target subsets are obtained, which have the characteristics of high similarity within subsets and the largest number of mutations, and high mutual exclusion among subsets and the largest number of subsets. This paper conducted a series of experiments to verify the performance of the algorithm, including simulation datasets and truthful datasets from TCGA. According to the results, the algorithm shows good performance under different simulation configurations, and some of the mutually exclusive patterns detected from TCGA datasets were supported by published literatures. This paper compared the performance to MEGSA, which is the best and most widely used method at present. The purities and computational efficiencies on simulation datasets outperformed MEGSA.

BACKGROUND:Genomic micro-satellites are the genomic regions that consist of short and repetitive DNA motifs. Estimating the length distribution and state of a micro-satellite region is an important computational step in cancer sequencing data pipelines, which is suggested to facilitate the downstream analysis and clinical decision supporting. Although several state-of-the-art approaches have been proposed to identify micro-satellite instability (MSI) events, they are limited in dealing with regions longer than one read length. Moreover, based on our best knowledge, all of these approaches imply a hypothesis that the tumor purity of the sequenced samples is sufficiently high, which is inconsistent with the reality, leading the inferred length distribution to dilute the data signal and introducing the false positive errors.RESULTS:In this article, we proposed a computational approach, named ELMSI, which detected MSI events based on the next generation sequencing technology. ELMSI can estimate the specific length distributions and states of micro-satellite regions from a mixed tumor sample paired with a control one. It first estimated the purity of the tumor sample based on the read counts of the filtered SNVs loci. Then, the algorithm identified the length distributions and the states of short micro-satellites by adding the Maximum Likelihood Estimation (MLE) step to the existing algorithm. After that, ELMSI continued to infer the length distributions of long micro-satellites by incorporating a simplified Expectation Maximization (EM) algorithm with central limit theorem, and then used statistical tests to output the states of these micro-satellites. Based on our experimental results, ELMSI was able to handle micro-satellites with lengths ranging from shorter than one read length to 10kbps.CONCLUSIONS:To verify the reliability of our algorithm, we first compared the ability of classifying the shorter micro-satellites from the mixed samples with the existing algorithm MSIsensor. Meanwhile, we varied the number of micro-satellite regions, the read length and the sequencing coverage to separately test the performance of ELMSI on estimating the longer ones from the mixed samples. ELMSI performed well on mixed samples, and thus ELMSI was of great value for improving the recognition effect of micro-satellite regions and supporting clinical decision supporting. The source codes have been uploaded and maintained at https://github.com/YixuanWang1120/ELMSI for academic use only.

2019年9月24日在四川省大竹县乌木滩水库进行水生生物多样性资源调查时,在乌木滩水库(30°44.269′N,107°17.029′E)采集到9尾鱼类标本,形态疑似太阳鱼.实验室采用DNA条形码技术,设计线粒体CO Ⅰ (Cytochrome oxidase subunit Ⅰ,COⅠ)基因通用引物(引物信息如表1所示),通过PCR扩增并测序,拼接得到长度为581bp的序列,在NCBI中进行BLAST比对,结果发现与数据库中蓝太阳鱼CO Ⅰ基因序列相似度高达99％,因此将此物种鉴定为蓝太阳鱼(Lepomis cyanellus),隶属于鲈形目(Perciformes)、棘臀鱼科(Centrarchidae)、太阳鱼属(Lepomis).

Background Genotyping of structural variation is an important computational problem in next generation sequence data analysis. However, in cancer genomes, the copy number variant(CNV) often coexists with other types of structural variations which significantly reduces the accuracy of the existing genotype methods. The bias on sequencing coverage and variant allelic frequency can be observed on a CNV region, which leads to the genotyping approaches that misinterpret the heterozygote as a homozygote. Furthermore, other data signals such as split mapped read, abnormal read will also be misjudged because of the CNV. Therefore, genotyping the structural variations with CNV is a complicated computational problem which should consider multiple features and their interactions. Methods Here we proposed a computational method for genotyping indels in the CNV region, which introduced a machine learning framework to comprehensively incorporate a set of data features and their interactions. We extracted fifteen kinds of classification features as input and different from the traditional genotyping problem, here the structure of variant may fall into types of normal homozygote, homozygous variant, heterozygous variant without CNV, heterozygous variant with a CNV on the mutated haplotype, and heterozygous variant with a CNV on the wild haplotype. The Multiclass Relevance Vector Machine (M-RVM) was used as a machine learning framework combined with the distribution characteristics of the features. Results We applied the proposed method to both simulated and real data, and compared it with the existing popular softwares include Gindel, Facets, GATK , and also compared with other machine learning cores: Support Vector Machine, Lanrange-SVM with OVO multiple classification, Naïve Bayes and BP Neural Network . The results demonstrated that the proposed method outperforms others on accuracy, stability and efficiency. Conclusion This work shows that the genotyping of structural variations on the CNV region cannot be solved as a traditional genotyping problem. More features should be used to efficiently complete the five-category task. According to the result, the proposed method can be a practical algorithm to correct genotype structural variations with CNV on the next generation sequence data. The source codes have been uploaded at https://github.com/TrinaZ/Mixgenotype for academic usage only.

Some somatic mutations are reported to present mutually exclusive patterns. It is a basic computational problem to efficiently extracting mutually exclusive patterns from cancer mutation data. In this article, we focus on the inter-set mutual exclusion problem, which is to group the genes into at least two sets, with the mutations in the different sets mutually exclusive. The proposed algorithm improves the calculation of the score of mutual exclusion. The improved measurement considers the percentage of supporting cases, the approximate exclusivity degree and the pair-wise similarities of two genes. Moreover, the proposed algorithm adopts a greedy strategy to generate the sets of genes. Different from the existing approaches, the greedy strategy considers the scores of mutual exclusion between both the genes and virtual genes, which benefits the selection with the size restrictions. We conducted a series of experiments to verify the performance on simulation datasets and TCGA dataset consisting of 477 real cases with more than 10 million mutations within 28507 genes. According to the results, our algorithm demonstrated good performance under different simulation configurations. In addition, it outperformed CoMEt, a widely-accepted algorithm, in recall rates and accuracies on simulation datasets. Moreover, some of the exclusive patterns detected from TCGA dataset were supported by published literatures.

OBJECTIVETo reconstruct tumor clonal haplotypes based on the third-generation sequencing data to effectively identify tumor heterogeneity.METHODSWe developed an algorithm for extracting somatic mutational event from the mixed tumor data and determining the connection weight of each somatic cell mutation site through the probability function. A reconstruction algorithm of the haplotype was designed based on the maximum spanning tree, and following the principle of inheritance between tumor clones, the connection pattern was determined at each mutation site in the clonal maximum spanning tree in a stepwise manner. The number, ratio and evolution of the sub-clones were estimated using the depth stripping method.RESULTSIn the simulation experiments, we analyzed the accuracy of the algorithm based on 4 indexes, namely the coverage, read length, subclone number and somatic variant rate, and the Results demonstrated a good robustness of the algorithm. The Results of the experiments showed that the mean sub-clone haplotypes accuracy exceeded 97%, suggesting that this algorithm significantly outperformed the previous Methods.CONCLUSIONSThe proposed method can accurately reconstruct tumor subclonal haplotypes and clarify the process of tumor clonal evolution, and can thus provide a theoretical basis for tumor heterogeneity research and assist in clinical decision-making.

采用形态学和多变量形态度量方法,对西伯利亚鲟(Acipenser baerii)、施氏鲟(Acipenser schrenckii)及其杂交种(西伯利亚鲟♀×施氏鲟♂)的形态异同进行了分析,以鉴别区分三者的形态特征.结果发现,西伯利亚鲟、施氏鲟及其杂交种的可数性状中鳃耙数和背鳍数均具有显著差异;可量性状的多重比较分析显示杂交种的眼间距/全长显著小于西伯利亚鲟和施氏鲟,三者的吻长/全长均具有显著差异;主成分分析提取的前三个主成分对变异的累积贡献率为65.68％;判别分析构建了西伯利亚鲟、施氏鲟及其杂交种的判别公式,判别公式预测分类总体准确率为85.6％.分析结果表明,西伯利亚鲟、施氏鲟及其杂交种间的形态差异主要体现在头部及尾柄.

Community structures are very common in complex networks. Detecting these communities is important for understanding the hidden features of networks. Besides, each community usually has one leader, which presents its significant influence over the whole community. However, most existing methods just focus on the problem of graph clustering, ignoring the role of community leaders. To solve this problem, in this paper, we propose a novel leader-aware community detection algorithm, which can find community structures as well as leaders of each community. This algorithm measures the leadership of each node and lets each one adhere to its local leader, forming dependence trees. Once all dependence trees are definitely settled, the community structures emerge because one tree actually is a cluster. Additionally, each root node of the tree is exactly the leader of corresponding community. This method can quickly determine the belonging of each node. Experimental results on real-world and benchmark networks demonstrate the effectiveness and the efficiency of our algorithm compared with other state-of-the-art approaches.

BackgroundHaplotype phasing is an important step in many bioinformatics workflows. In cancer genomics, it is suggested that reconstructing the clonal haplotypes of a tumor sample could facilitate a comprehensive understanding of its clonal architecture and further provide valuable reference in clinical diagnosis and treatment. However, the sequencing data is an admixture of reads sampled from different clonal haplotypes, which complicates the computational problem by exponentially increasing the solution-space and leads the existing algorithms to an unacceptable time-/space- complexity. In addition, the evolutionary process among clonal haplotypes further weakens those algorithms by bringing indistinguishable candidate solutions.ResultsTo improve the algorithmic performance of phasing clonal haplotypes, in this article, we propose MixSubHap, which is a graph-based computational pipeline working on cancer sequencing data. To reduce the computation complexity, MixSubHap adopts three bounding strategies to limit the solution space and filter out false positive candidates. It first estimates the global clonal structure by clustering the variant allelic frequencies on sampled point mutations. This offers a priori on the number of clonal haplotypes when copy-number variations are not considered. Then, it utilizes a greedy extension algorithm to approximately find the longest linkage of the locally assembled contigs. Finally, it incorporates a read-depth stripping algorithm to filter out false linkages according to the posterior estimation of tumor purity and the estimated percentage of each sub-clone in the sample. A series of experiments are conducted to verify the performance of the proposed pipeline.ConclusionsThe results demonstrate that MixSubHap is able to identify about 90% on average of the preset clonal haplotypes under different simulation configurations. Especially, MixSubHap is robust when decreasing the mutation rates, in which cases the longest assembled contig could reach to 10kbps, while the accuracy of assigning a mutation to its haplotype still keeps more than 60% on average. MixSubHap is considered as a practical algorithm to reconstruct clonal haplotypes from cancer sequencing data. The source codes have been uploaded and maintained at https://github.com/YixuanWang1120/MixSubHap for academic use only.

Most of the repetitive elements in the human genome are associated with retrotransposons, which have wide-ranging impacts on complex traits and diseases. Detecting human active transposon LINE-1 insertions is a tricky computational problem because of their repetitiveness and similarities. Existing methods are not working well for identifying large-scale insertion events, or rely on a small number of annotated samples, which often leads to high false positive rates. In this paper, we proposed a semi-supervised framework, named L1Detector, to improve the performance of the detection and classification processes. The core of L1Detector was a shallow neural network. This framework first extracted multiple features around the candidate insertion sites. Then, it took the advantages of an existing machine learning model to compute the interactions among the features. We further improved this model by introducing a semi-supervised learning framework, which facilitated to handle the large-scale unlabeled data. In addition, this framework enhanced a comprehensively and accurately detection on the polymorphic insertion events and insertion types. We conducted a series of simulation experiments to evaluate the performance of the proposed framework and compared it to a popular detection method. The experiment results demonstrated that the proposed framework often provided more comprehensive and effective results.

为了解西伯利亚鲟(Acipenser baeri)骨骼系统的形态特征,采用传统的解剖法和透明骨法对西伯利亚鲟幼鱼进行解剖观察,为西伯利亚鲟在形态学和分类学的研究提供基础资料.结果表明:西伯利亚鲟幼鱼骨骼系统由主轴骨骼和附肢骨骼两部分构成,主轴骨骼包括头骨、脊柱和肋骨,附肢骨骼由鳍条、支鳍骨骼和带骨构成.对比分析发现,西伯利亚鲟和施氏鲟骨骼系统的组成与构造较为相似,其原因可能是其对高纬度低水温环境长期适应的结果.

Abstract Complex insertion and deletion (complex indel) is a rare category of genomic structural variations, which is formed by inserting one or multiple DNA fragments into the genomic location where a deletion occurs. A recent study conducts a systematic analysis from over 8,000 pan-cancer cases, which reports hundreds of complex indels in cancer-associated genes, some of which are considered potentially druggable. Several approaches are proposed to detect complex indels from the new-generation sequencing data (both 2nd and 3rd generation). However, as different data-mining algorithms vary the preferences on capturing data patterns, different approaches may report the conflicted complex indel calls. Here, we propose a machine learning approach to correct the conflicted calls from different approaches and further estimate the genotype of each call. The proposed approach, implemented as CIMerge, adopts a relevance vector machine framework. For each candidate call, CIMerge first extracts a set of features on the candidate region, which includes the read depth, variant allelic frequency, number of the splitting/unmapped reads, number of the discordant paired-end reads, aligned contig, etc. CIMerge also considers another set of features on data-mining algorithm(s) that reported the candidate call, which includes the parameter settings, etc. Both sets of features are trained by the relevance vector machine framework, which outputs the probability of each candidate of the conflicted call. As a byproduct, it outputs the genotype of the candidate call with highest likelihood. We tested CIMerge on multiple datasets generated by different simulation configurations and compared it to several state-of-the-art approaches. The experiment results demonstrate that CIMerge outperforms the existing approaches. The average success rate of recognition is approaching 90%, while Pindel and Gindel are reported as 62.53% and 65.202%, respectively. The software package CIMerge is freely available for academic uses at https://github.com/xjtu712-lab/CIMerge. Citation Format: Tian Zheng, Yang Li, Yu Geng, Zhongmeng Zhao, Xuanping Zhang, Jiayin Wang. CIMerge: A machine learning approach for merging and genotyping complex indel calls from NGS data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 5294.

Along with tumor growth, somatic alternations are continually accumulating, some of which leads to the formations of clonal populations. Genomic deletion is a major type of such genomic alternations. Although tens of computational methods were published, in the past decade, for detecting genomic deletions from next generation sequencing data, the existing algorithms often suffer an accuracy loss when they encounter the cases of deletion calls with complex boundaries. It is reported that a genomic deletion that occurs in different sub-clones may present nearby boundaries. Such deletion is considered as a deletion with complex boundaries. The existing approaches either ignore the complex-boundary cases by reporting the pair of boundaries with the largest numbers of supporting reads, or even provide incorrect results due to the interference data signals. To overcome this weakness, in this paper, we propose a heuristic method, SV-Del, to help the popular methods correct the detection errors, which are introduced by complex boundaries. The results of an existing method are the given candidate calls. SV-Del filters these calls and identifies the ones with complex boundaries. The proposed method first adopts a segmented extension algorithm and utilizes the longest variable splitting-read strategy to detect the possible pairs of boundaries in each candidate region. Then, it uses the longest variable splitting-reads to correct the detection errors which may introduced by clonal SNVs. To differentiate the detection errors from possible pairs of deletion boundaries, SV-Del estimates the numbers of sub-clones across sampled candidate regions, and then it uses a gradually separating algorithm to attain and refine the candidate calls. We applied SV-Del on a series of simulated datasets which are generated by different settings. The experiment results demonstrate that the detection accuracy is significantly improved comparing to the original results. SV-Del is also shown robust. The source codes and software package of SV-Del are uploaded at https://github.com/Hope523/SV-Del for academic uses only.

Complex insertion and deletion (complex indel) is a rare category of genomic structural variations, which is formed by inserting one or multiple DNA fragments into the genomic location where a deletion occurs. A recent study conducts a systematic analysis from over 8,000 pan-cancer cases, which reports hundreds of complex indels in cancer-associated genes, some of which are considered potentially druggable. Several approaches are proposed to detect complex indels from the new-generation sequencing data (both 2 nd and 3 rd generation). However, as different data-mining algorithms vary the preferences on capturing data patterns, different approaches may report the conflicted complex indel calls. Here, we propose a machine learning approach to correct the conflicted calls from different approaches and further estimate the genotype of each call. The proposed approach, implemented as CIMerge, adopts a relevance vector machine framework. For each candidate call, CIMerge first extracts a set of features on the candidate region, which includes the read depth, variant allelic frequency, number of the splitting/unmapped reads, number of the discordant paired-end reads, aligned contig, etc. CIMerge also considers another set of features on data-mining algorithm(s) that reported the candidate call, which includes the parameter settings, etc. Both sets of features are trained by the relevance vector machine framework, which outputs the probability of each candidate of the conflicted call. As a byproduct, it outputs the genotype of the candidate call with highest likelihood. We tested CIMerge on multiple datasets generated by different simulation configurations and compared it to several state-of-the-art approaches. The experiment results demonstrate that CIMerge outperforms the existing approaches. The average success rate of recognition is approaching 90%, while Pindel and Gindel are reported as 62.53% and 65.202%, respectively. The software package CIMerge is freely available for academic uses at https://github.com/xjtu712-lab/CIMerge. Citation Format: Tian Zheng, Yang Li, Yu Geng, Zhongmeng Zhao, Xuanping Zhang, Jiayin Wang. CIMerge: A machine learning approach for merging and genotyping complex indel calls from NGS data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 5294.

In recent years, the next generation sequencing enables us to obtain high resolution landscapes of the genetic changes at single-nucleotide level. More and more novel methods are proposed for efficient and effective analyses on cancer sequencing data. To facilitate such development, data simulator is a crucial tool, which not only tests and evaluates proposed approaches, but provides the feedbacks for further improvements as well. Several simulators are released to generate the next generation sequencing data. However, based on our best knowledge, none of them considers clonality information. It is suggested that clonal heterogeneity does widely exist in tumor samples. The patterns of somatic mutational events usually expose a wide spectrum of variant allelic frequencies, while some of them are only detectable in one or multiple clonal lineages. In this article, we introduce a Tumor-Normal sequencing Simulator, TNSim, to generate the next generation sequencing data by involving clonality information. The simulator is able to mimic a tumor sample and the paired normal sample, where the germline variants and somatic mutations can be settled respectively. Tumor purity is adjustable. Clonal architecture is preassigned as one or more clonal lineages, where each lineage consists of a set of somatic mutations whose variant allelic frequencies are similar. A group of experiments are conducted to evaluate its performance. The statistical features of the artificial sequencing reads are comparable to the real tumor sequencing data whose sample consists of multiple sub-clones. The source codes are available at http://github.com/lnmxgy/TNSim and for academic use only.

In social network analysis, community detection on plain graphs has been widely studied. With the proliferation of available data, each user in the network is usually associated with additional attributes for elaborate description. However, many existing methods only concentrate on the topological structure and fail to deal with node‐attributed networks. These approaches are incapable of extracting clear semantic meanings for communities detected. In this paper, we combine the topological structure and attribute information into a unified process and propose a novel algorithm to detect overlapping semantic communities. Moreover, a new metric is designed to measure the density of semantic communities. The proposed algorithm is divided into 3 phases. First, we detect local semantic subcommunities from each node's perspective using a greedy strategy on the metric. Then, a supergraph, which consists of all these subcommunities is created. Finally, we find global semantic communities on the supergraph. The experimental results on real‐world data sets show the efficiency and effectiveness of our approach against other state‐of‐the‐art methods.

The association analysis between SNP and disease phenotype is of great significance in medical diagnosis and treatment. However, the existing research methods are for a single disease or phenotype, which is a "multiple-loci-single phenotype" situation. It is not suitable for practical conditions which are often "multiple-loci-multiple-phenotypes", and the situation does not take into account the relationship between the internals of the phenotypes. If the existing methods are used to achieve the above purpose, the same processing needs to be done once for each phenotype resulting huge resource cost. When mining the logical relationship between phenotypes, the combination is exponentially increasing. Therefore, its time complexity is also exponentially increasing. In this paper, a method based on particle swarm optimization and hierarchical clustering is proposed. It consists of 4 steps. Firstly, the phenotypic dataset are clustered and then combined, after that the algorithm model is initialized, and the particle swarm optimization (PSO) algorithm is used for extended optimization, and finally the result is filtered to obtain the final result. In this paper, the experiment is carried out by controlling the total number of SNP loci, the number of pathogenic mutation loci to be excavated, the total number of phenotypes in the phenotypic dataset, and the number of phenotypes to be mined. For every configuration of different parameter values, 100 sets of data were simulated for experimentation. The results show that the proposed method adapts to the "multiple-loci-multiple-phenotypes" case and obtained highly accurate results. When performing the algorithm, it does not need to loop the routine for each phenotype. The most probable candidates are pre-selected according to the fitness function of PSO algorithm. Therefore, the time and system resource costs are largely reduced comparing with previous methods. The source code and testing datasets are uploaded at https://github.com/GALI17/PSO.

Complex insertion and deletion (complex indel) is a rare category of genomic structural variations, which is formed by inserting one or multiple DNA fragments into the genomic location where a deletion occurs. A recent study conducts a systematic analysis from over 8,000 pan-cancer cases, which reports hundreds of complex indels in cancer-associated genes, some of which are considered potentially druggable. Several approaches are proposed to detect complex indels from the new-generation sequencing data (both 2 nd and 3 rd generation). However, as different data-mining algorithms vary the preferences on capturing data patterns, different approaches may report the conflicted complex indel calls. Here, we propose a machine learning approach to correct the conflicted calls from different approaches and further estimate the genotype of each call. The proposed approach, implemented as CIMerge, adopts a relevance vector machine framework. For each candidate call, CIMerge first extracts a set of features on the candidate region, which includes the read depth, variant allelic frequency, number of the splitting/unmapped reads, number of the discordant paired-end reads, aligned contig, etc. CIMerge also considers another set of features on data-mining algorithm(s) that reported the candidate call, which includes the parameter settings, etc. Both sets of features are trained by the relevance vector machine framework, which outputs the probability of each candidate of the conflicted call. As a byproduct, it outputs the genotype of the candidate call with highest likelihood. We tested CIMerge on multiple datasets generated by different simulation configurations and compared it to several state-of-the-art approaches. The experiment results demonstrate that CIMerge outperforms the existing approaches. The average success rate of recognition is approaching 90%, while Pindel and Gindel are reported as 62.53% and 65.202%, respectively. The software package CIMerge is freely available for academic uses at https://github.com/xjtu712-lab/CIMerge. Citation Format: Tian Zheng, Yang Li, Yu Geng, Zhongmeng Zhao, Xuanping Zhang, Jiayin Wang. CIMerge: A machine learning approach for merging and genotyping complex indel calls from NGS data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 5294.

This article has been withdrawn at the request of the authors and the Editor. The Publisher apologizes for any inconvenience this may cause. The full Elsevier Policy on Article Withdrawal can be found at http://www.elsevier.com/locate/withdrawalpolicy.

Genomic micro-satellites are the genomic regions that consist of short and repetitive DNA motifs. In contrast to unique genome, genomic micro-satellites expose high intrinsic polymorphisms, which mainly derive from variability in length. Length distributions are widely used to represent the polymorphisms. Recent studies report that some micro-satellites alter their length distributions significantly in tumor tissue samples comparing to the ones observed in normal samples, which becomes a hot topic in cancer genomics. Several state-of-the-art approaches are proposed to identify the length distributions from the sequencing data. However, the existing approaches can only handle the micro-satellites shorter than one read length, which limits the potential research on long micro-satellite events. In this article, we propose a probabilistic approach, implemented as ELMSI that estimates the length distributions of the micro-satellites longer than one read length. The core algorithm works on a set of mapped reads. It first clusters the reads, and a k-mer extension algorithm is adopted to detect the unit and breakpoints as well. Then, it conducts an expectation maximization algorithm to approach the true length distributions. According to the experiments, ELMSI is able to handle micro-satellites with the length spectrum from shorter than one read length to 10 kbps scale. A series of comparison experiments are applied, which vary the numbers of micro-satellite regions, read lengths and sequencing coverages, and ELMSI outperforms MSIsensor in most of the cases.