Shibin Qiu

A computational study of off-target effects of RNA interference

RNA interference (RNAi) is an intracellular mechanism for post-transcriptional gene silencing that is frequently used to study gene function. RNAi is initiated by short interfering RNA (siRNA) of ∼21 nt in length, either generated from the double-stranded RNA (dsRNA) by using the enzyme Dicer or introduced experimentally. Following association with an RNAi silencing complex, siRNA targets mRNA transcripts that have sequence identity for destruction. A phenotype resulting from this knockdown of expression may inform about the function of the targeted gene. However, ‘off-target effects’ compromise the specificity of RNAi if sequence identity between siRNA and random mRNA transcripts causes RNAi to knockdown expression of non-targeted genes. The complete off-target effects must be investigated systematically on each gene in a genome by adjusting a group of parameters, which is too expensive to conduct experimentally and motivates a study in silico. This computational study examined the potential for off-target effects of RNAi, employing the genome and transcriptome sequence data of Homo sapiens, Caenorhabditis elegans and Schizosaccharomyces pombe. The chance for RNAi off-target effects proved considerable, ranging from 5 to 80% for each of the organisms, when using as parameter the exact identity between any possible siRNA sequences (arbitrary length ranging from 17 to 28 nt) derived from a dsRNA (range 100–400 nt) representing the coding sequences of target genes and all other siRNAs within the genome. Remarkably, high-sequence specificity and low probability for off-target reactivity were optimally balanced for siRNA of 21 nt, the length observed mostly in vivo. The chance for off-target RNAi increased (although not always significantly) with greater length of the initial dsRNA sequence, inclusion into the analysis of available untranslated region sequences and allowing for mismatches between siRNA and target sequences. siRNA sequences from within 100 nt of the 5′ termini of coding sequences had low chances for off-target reactivity. This may be owing to coding constraints for signal peptide-encoding regions of genes relative to regions that encode for mature proteins. Off-target distribution varied along the chromosomes of C.elegans, apparently owing to the use of more unique sequences in gene-dense regions. Finally, biological and thermodynamical descriptors of effective siRNA reduced the number of potential siRNAs compared with those identified by sequence identity alone, but off-target RNAi remained likely, with an off-target error rate of ∼10%. These results also suggest a direction for future in vivo studies that could both help in calibrating true off-target rates in living organisms and also in contributing evidence toward the debate of whether siRNA efficacy is correlated with, or independent of, the target molecule. In summary, off-target effects present a real but not prohibitive concern that should be considered for RNAi experiments.

A Framework for Multiple Kernel Support Vector Regression and Its Applications to siRNA Efficacy Prediction

The cell defense mechanism of RNA interference has applications in gene function analysis and promising potentials in human disease therapy. To effectively silence a target gene, it is desirable to select appropriate initiator siRNA molecules having satisfactory silencing capabilities. Computational prediction for silencing efficacy of siRNAs can assist this screening process before using them in biological experiments. String kernel functions, which operate directly on the string objects representing siRNAs and target mRNAs, have been applied to support vector regression for the prediction and improved accuracy over numerical kernels in multidimensional vector spaces constructed from descriptors of siRNA design rules. To fully utilize information provided by string and numerical data, we propose to unify the two in a kernel feature space by devising a multiple kernel regression framework where a linear combination of the kernels is used. We formulate the multiple kernel learning into a quadratically constrained quadratic programming (QCQP) problem, which although yields global optimal solution, is computationally demanding and requires a commercial solver package. We further propose three heuristics based on the principle of kernel-target alignment and predictive accuracy. Empirical results demonstrate that multiple kernel regression can improve accuracy, decrease model complexity by reducing the number of support vectors, and speed up computational performance dramatically. In addition, multiple kernel regression evaluates the importance of constituent kernels, which for the siRNA efficacy prediction problem, compares the relative significance of the design rules. Finally, we give insights into the multiple kernel regression mechanism and point out possible extensions.

Discrete dynamic Bayesian network analysis of fMRI data

We examine the efficacy of using discrete Dynamic Bayesian Networks (dDBNs), a data‐driven modeling technique employed in machine learning, to identify functional correlations among neuroanatomical regions of interest. Unlike many neuroimaging analysis techniques, this method is not limited by linear and/or Gaussian noise assumptions. It achieves this by modeling the time series of neuroanatomical regions as discrete, as opposed to continuous, random variables with multinomial distributions. We demonstrated this method using an fMRI dataset collected from healthy and demented elderly subjects (Buckner, et al., 2000 : J Cogn Neurosci 12:24‐34) and identify correlates based on a diagnosis of dementia. The results are validated in three ways. First, the elicited correlates are shown to be robust over leave‐one‐out cross‐validation and, via a Fourier bootstrapping method, that they were not likely due to random chance. Second, the dDBNs identified correlates that would be expected given the experimental paradigm. Third, the dDBNs ability to predict dementia is competitive with two commonly employed machine‐learning classifiers: the support vector machine and the Gaussian naïve Bayesian network. We also verify that the dDBN selects correlates based on non‐linear criteria. Finally, we provide a brief analysis of the correlates elicited from Buckner et al.s data that suggests that demented elderly subjects have reduced involvement of entorhinal and occipital cortex and greater involvement of the parietal lobe and amygdala in brain activity compared with healthy elderly (as measured via functional correlations among BOLD measurements). Limitations and extensions to the dDBN method are discussed. Hum Brain Mapp, 2009.

Multiple kernel support vector regression for siRNA efficacy prediction

The cell defense mechanism of RNA interference has applicationsin gene function analysis and human disease therapy. To effectivelysilence a target gene, it is desirable to select the initiator siRNA moleculeshaving satisfactory silencing capabilities. Computational prediction forsilencing efficacy of siRNAs can assist this screening process before usingthem in biological experiments. String kernel functions, which operatedirectly on the string objects representing siRNAs and target mRNAs,have been applied to support vector regression for the prediction and improvedaccuracy over numerical kernels in multidimensional vector spacesconstructed from descriptors of siRNA design rules. To fully utilize informationprovided by string and numerical kernels, we propose to unifythe two in the kernel feature space by devising a multiple kernel regressionframework where a linear combination of the kernels are used. Weformulate the multiple kernel learning into a quadratically constrainedquadratic programming (QCQP) problem, which although yields globaloptimal solution, is computationally inefficient and requires a commercialsolver package.We further propose three heuristics based on the principleof kernel-target alignment and predictive accuracy. Empirical results onreal biological data demonstrate that multiple kernel regression can improveaccuracy and decrease model complexity by reducing the numberof support vectors. In addition, multiple kernel regression gives insightsinto the kernel combination, which, for siRNA efficacy prediction, evaluatesthe relative significance of the design rules.

The RNA String Kernel for siRNA Efficacy Prediction

String kernels directly model sequence similarities without the necessity of extracting numerical features in a vector space. Since they better capture complex traits in the sequences, string kernels often achieve better prediction performance. RNA interference is a cell defense mechanism with many biological and therapeutical applications, where strings can be used to represent target messenger RNAs and initiating short RNAs and string kernels can be applied for training and prediction. While most existing string kernels are developed for general purpose sequences and have been applied to text and protein classifications, the RNA string kernel is particularly designed to model mismatches, GU wobbles, and bulges of RNA biology and has been applied to RNAi off-target evaluation. We adapt the RNA string kernel to compute the similarity of siRNA sequences and use it in support vector regression to predict siRNA silencing efficacy. We evaluate the performance of the RNA kernel against the spectrum kernel, the string subsequence kernel of arbitrary mismatch, the randomized string kernel, and numerical kernels computed from numerical features extracted according to siRNA design rules. We also give insights into computational performance and common properties and differences of the RNA kernel as compared to other kernels. Empirical results on biological data sets demonstrate that the RNA string kernel performed favorably than most existing string kernels and achieved significant improvements over kernels computed from numerical descriptors extracted according to structural and thermodynamic rules. Meanwhile, the string kernels achieved favorable results relative to other methods in related work. Furthermore, the RNA string kernel is simple to implement and fast to compute.

String kernels of imperfect matches for off-target detection in RNA interference

RNA interference (RNAi) is a posttranscriptional gene silencing mechanism frequently used to study gene functions and knock down viral genes. RNAi has been regarded as a highly effective means of gene repression. However, an “off-target effect” deteriorates its specificity and applicability. The complete off-target effects can only be characterized by examining all factors through systematic investigation of each gene in a genome. However, this complete investigation is too expensive to conduct experimentally which motivates a computational study. The sequence matching between an siRNA and its target mRNA allows for mismatches, G-U wobbles, and the secondary structure bulges, in addition to exact matches. To simulate these matching features, we propose string kernels measuring the similarity between two oligonucleotides and develop novel efficient implementations for RNAi off-target detection. We apply the algorithms for off-target errors in C. elegans and human.

Efficient search algorithms for RNAi target detection

Abstract RNA interference (RNAi) is a posttranscriptional gene silencing mechanism used to study gene functions, inhibit viral activities, and treat diseases therapeutically. However, RNAi has off-target effects—non-target genes can be unintentionally silenced. Therefore, target validation through target detection is crucial for the success of RNAi experiments. Effective target detection must examine each gene expressed by an organism, making computational efficiency a critical issue. We develop efficient sequential and parallel search algorithms using RNA string kernels, which model mismatches, G-U wobbles, bulges, and the seed region in the hybridization between an siRNA and its target mRNA. Empirical results demonstrate that our algorithms achieved speedups of six orders of magnitude over the alignment algorithm based on tests in the organisms of S. pombe, C. elegans, D. melanogaster, and human. Our design strategy also leads to a framework for efficient, flexible, and portable string search algorithms allowing for inexact matches.

Phase transitions in gene knockdown networks of transitive RNAi

Since gene silencing by RNA interference (RNAi) has been observed when inexact matches exist in siRNA-mRNA binding, the number of mismatched nucleotides allowed by nature becomes an important quantity in characterizing RNAi specificity. We use scale-free graphs to model the knockdown interactions among different genes and estimate the allowable flexibility by examining transitive RNAi, which amplifies siRNA and cyclically silences targets. Simulation results in S. pombe indicate that continually increasing the number of mismatches risks transcriptome-wide knockdown and eventually turns RNAi from defensive to self-destructive. At the phase transition, the number of mismatches indicates a critical value beyond which tRNAi would cause an organism instable. This critical value suggests an upper limit of no more than 6 nt mismatches in the binding.

SiRNA silencing efficacy prediction using the RNA string kernel.

While most existing string kernels are developed for general purpose sequences and have been applied to text and protein classifications, the RNA string kernel is particularly designed to model mismatches, G-U wobbles, and bulges of RNA biology. We adapt the RNA kernel to compute the similarity of the short interfering RNAs (siRNAs), initiators of RNA interference, and use it in support vector regression to predict the siRNA silencing efficacy treated as a continuous variable. Empirical results on biological data sets demonstrate that the RNA string kernel performed favourably. In addition, it is simple to implement and fast to compute.

Implications of Phase Transitions in Knockdown Networks of Transitive RNAi

Gene silencing by RNA interference (RNAi) has been observed even in the presence of imperfect complementarity in the siRNA-mRNA hybridization. Since more permissive mismatches gives rise to higher chances of off-target gene silencing, the number of mismatched nucleotides allowed by nature becomes an important quantity in characterizing RNAi specificity and RNAi design. To estimate the allowable flexibility, we use scale-free graphs to model the knockdown interactions among genes by examining transitive RNAi (tRNAi), which amplifies siRNA and cyclically silences targets. We removed inefficient siRNA sequences using the commonly used siRNA efficacy rules, avoided redundant siRNAs using barcoding techniques, and employed both contiguous and scattered mismatches to emulate the siRNA-mRNA binding. Simulations in multiple organisms indicate that the fraction of the transcriptome silenced by tRNAi rises drastically with increased number of allowed mismatches and eventually tRNAi became self-destructive rather than defensive. At the phase transition, the number of mismatches implies a critical value beyond which tRNAi would cause the transcription of an organism to be instable. This critical value suggests an upper limit of no more than 6 nt mismatches in the hybridization in general