The paper introduces a method to identify slow molecular order parameters for constructing Markov models. The goal is to characterize slow relaxation processes in macromolecules by identifying structural changes and estimating their timescales. Traditional approaches, such as Markov models and Master equation models, discretize the state space and use eigenvectors and eigenvalues of transition matrices. However, identifying slow order parameters in high-dimensional spaces is challenging without prior knowledge. The authors propose using the variational principle of conformation dynamics to derive an optimal way to identify the "slow subspace" of a set of order parameters. This method is shown to be equivalent to time-lagged independent component analysis (TICA), a statistical technique that combines covariance and time-lagged covariance matrices. TICA is used to identify optimal indicators—order parameters that reflect slow transitions and can serve as reaction coordinates. The method is demonstrated on two molecular systems: the 6-residue fluorescent peptide MR121-GSGSW and the 30-residue natively disordered peptide KID. The slow subspace is shown to be effective in constructing accurate kinetic models of these systems. The identified optimal indicators reveal the structural changes associated with slow processes. The paper also discusses the theory behind the method, including the variational principle, the approximation of eigenfunctions, and the relationship between TICA timescales and Markov model timescales. It shows that TICA timescales are often underestimated, but Markov models can provide better estimates by discretizing the TICA subspace. The results demonstrate that TICA coordinates are useful for constructing Markov models, particularly for systems with complex conformational dynamics. The method is validated using simulations and benchmark data, showing its effectiveness in identifying slow processes and their associated timescales.The paper introduces a method to identify slow molecular order parameters for constructing Markov models. The goal is to characterize slow relaxation processes in macromolecules by identifying structural changes and estimating their timescales. Traditional approaches, such as Markov models and Master equation models, discretize the state space and use eigenvectors and eigenvalues of transition matrices. However, identifying slow order parameters in high-dimensional spaces is challenging without prior knowledge. The authors propose using the variational principle of conformation dynamics to derive an optimal way to identify the "slow subspace" of a set of order parameters. This method is shown to be equivalent to time-lagged independent component analysis (TICA), a statistical technique that combines covariance and time-lagged covariance matrices. TICA is used to identify optimal indicators—order parameters that reflect slow transitions and can serve as reaction coordinates. The method is demonstrated on two molecular systems: the 6-residue fluorescent peptide MR121-GSGSW and the 30-residue natively disordered peptide KID. The slow subspace is shown to be effective in constructing accurate kinetic models of these systems. The identified optimal indicators reveal the structural changes associated with slow processes. The paper also discusses the theory behind the method, including the variational principle, the approximation of eigenfunctions, and the relationship between TICA timescales and Markov model timescales. It shows that TICA timescales are often underestimated, but Markov models can provide better estimates by discretizing the TICA subspace. The results demonstrate that TICA coordinates are useful for constructing Markov models, particularly for systems with complex conformational dynamics. The method is validated using simulations and benchmark data, showing its effectiveness in identifying slow processes and their associated timescales.